Cover Story An Algebraic Method for Super Resolution Image Reconstruction 5. Technical Trends Applications of Image Processing in Industries 8

Transcription

1 ISSN X Volume No. 36 Issue No. 4 July 2012 ` 50/- Article Image and Video Processing Toolbox in Scilab 20 Research Front Accurate Pupil and Iris Localization using Reverse Function 16 Cover Story An Algebraic Method for Super Resolution Image Reconstruction 5 Technical Trends Applications of Image Processing in Industries 8 Research Front Some Upcoming Challenges in Bioimage Informatics 12 Article Importance of Shifting Focus in Solving Problems 23 CSI Communications ns July A

2

3 CSI Communications Contents Volume No. 36 Issue No. 4 July 2012 Editorial Board Chief Editor Dr. R M Sonar Editors Dr. Debasish Jana Dr. Achuthsankar Nair Resident Editor Mrs. Jayshree Dhere Advisors Dr. T V Gopal Mr. H R Mohan Published by Executive Secretary Mr. Suchit Gogwekar For Computer Society of India Design, Print and Dispatch by CyberMedia Services Limited Please note: CSI Communications is published by Computer Society of India, a non-profit organization. Views and opinions expressed in the CSI Communications are those of individual authors, contributors and advertisers and they may differ from policies and official statements of CSI. These should not be construed as legal or professional advice. The CSI, the publisher, the editors and the contributors are not responsible for any decisions taken by readers on the basis of these views and opinions. Although every care is being taken to ensure genuineness of the writings in this publication, CSI Communications does not attest to the originality of the respective authors content CSI. All rights reserved. Instructors are permitted to photocopy isolated articles for non-commercial classroom use without fee. For any other copying, reprint or republication, permission must be obtained in writing from the Society. Copying for other than personal use or internal reference, or of articles or columns not owned by the Society without explicit permission of the Society or the copyright owner is strictly prohibited PLUS Cover Story An Algebraic Method for Super Resolution Image Reconstruction Bhabatosh Chanda Technical Trends Applications of Image Processing in Industries Dr. Tanushyam Chattopadhyay, Brojeshwar Bhowmick, and Aniruddha Sinha Research Front Some Upcoming Challenges in Bioimage Informatics Saurav Basu Accurate Pupil and Iris Localization using Reverse Function R P Ramkumar and Dr. S Arumugam Articles Image and Video Processing Toolbox in Scilab Hema Ramachandran Importance of Shifting Focus in Solving Problems Dr. Pramod Koparkar Distance Units in CSS3 Hareesh N Nampoothiri Practitioner Workbench Programming.Tips()» Multidimensional Plots in Matlab for Data Analysis Baisa L Gunjal and Dr. Suresh N Mali Programming.Learn ( Python )» Read and Write Using Python Umesh P Security Corner Information Security» Privacy & Responsibility Adv. Prashant Mali IT Act 2000» Prof. IT Law in Conversation with Mr. IT Executive: Issue No. 4 Mr. Subramaniam Vutha ICT@ Society Upload India, Upload Achuthsankar S Nair Brain Teaser Dr. Debasish Jana 33 Ask an Expert Dr. Debasish Jana 34 Happenings@ICT: ICT News Briefs in June 2012 H R Mohan 35 CSI Report Dr. Dharm Singh 36 CSI News 37 Published by Suchit Gogwekar for Computer Society of India at Unit No. 3, 4th Floor, Samruddhi Venture Park, MIDC, Andheri (E), Mumbai Tel. : Fax : hq@csi-india.org Printed at GP Offset Pvt. Ltd., Mumbai CSI Communications July

4 Know Your CSI Executive Committee ( /14)» President Vice-President Hon. Secretary Mr. Satish Babu Prof. S V Raghavan Mr. S Ramanathan president@csi-india.org vp@csi-india.org secretary@csi-india.org Hon. Treasurer Immd. Past President Mr. V L Mehta Mr. M D Agrawal treasurer@csi-india.org ipp@csi-india.org Nomination Committee ( ) Dr. D D Sarma Mr. Bipin V Mehta Mr. Subimal Kundu Regional Vice-Presidents Region - I Region - II Region - III Region - IV Mr. R K Vyas Prof. Dipti Prasad Mukherjee Mr. Anil Srivastava Mr. Sanjeev Kumar Delhi, Punjab, Haryana, Himachal Assam, Bihar, West Bengal, Gujarat, Madhya Pradesh, Jharkhand, Chattisgarh, Pradesh, Jammu & Kashmir, North Eastern States Rajasthan and other areas Orissa and other areas in Uttar Pradesh, Uttaranchal and and other areas in in Western India Central & South other areas in Northern India. East & North East India rvp3@csi-india.org Eastern India rvp1@csi-india.org rvp2@csi-india.org rvp4@csi-india.org Region - V Region - VI Region - VII Region - VIII Prof. D B V Sarma Mr. C G Sahasrabudhe Mr. Ramasamy S Mr. Pramit Makoday Karnataka and Andhra Pradesh Maharashtra and Goa Tamil Nadu, Pondicherry, International Members rvp5@csi-india.org rvp6@csi-india.org Andaman and Nicobar, rvp8@csi-india.org Kerala, Lakshadweep rvp7@csi-india.org Division Chairpersons, National Student Coordinator & Publication Committee Chairman Division-I : Hardware ( ) Division-II : Software ( ) Division-III : Applications ( ) National Student Coordinator Dr. C R Chakravarthy Dr. T V Gopal Dr. Debesh Das Mr. Ranga Raj Gopal div1@csi-india.org div2@csi-india.org div3@csi-india.org Division-IV : Communications Division-V : Education and Research Publication Committee ( ) ( ) Chairman Mr. Sanjay Mohapatra Chairman Division V Prof. R K Shyamsundar div4@csi-india.org To be announced div5@csi-india.org Important links on CSI website» Structure & Organisation National, Regional & State Students Coordinators Statutory Committees Collaborations Join Now - Renew Membership Member Eligibility Member Benefits Subscription Fees Forms Download BABA Scheme Publications CSI Communications* Adhyayan* R & D Projects Technical Papers Tutorials Course Curriculum Training Program (CSI Education Products) Travel support for International Conference enewsletter* Current Issue Archives Policy Guidelines Events President s Desk * Access is for CSI members only. Important Contact Details» For queries, correspondence regarding Membership, contact helpdesk@csi-india.org ExecCom Transacts News & Announcements archive CSI Divisions and their respective web links Division-Hardware Division Software Division Application Division Communications Division Education and Research List of SIGs and their respective web links SIG-Artificial Intelligence SIG-eGovernance SIG-FOSS SIG-Software Engineering SIG-DATA SIG-Distributed Systems SIG-Humane Computing SIG-Information Security SIG-Web 2.0 and SNS SIG-BVIT SIG-WNs SIG-Green IT SIG-HPC SIG-TSSR Other Links - Forums Blogs Communities* CSI Chapters Calendar of Events CSI Communications July

5 President s Message Satish Babu From : president@csi-india.org Subject : President s Desk Date : 1st July, 2012 Dear Members The ExeCom of CSI met on 30th June, 2012 and reviewed the activities and plans of CSI. Based on the recommendations of the Nominations Committee, the ExeCom also filled an interim vacancy of Chair, Division V, with Prof. RP Soni, an eminent professional from Ahmedabad. I would like to welcome Prof. Soni to the CSI ExeCom on behalf of all of us. International Activities One of the major areas of discussion during the June ExeCom was international activities and linkages of CSI. These constitute an important role of CSI in fulfillment of its position as a national society. I would like to update you briefly on these. IFIP IFIP and SEARCC are two international agencies that CSI, as a national society, are directly involved with. CSI is a member of IFIP, which was founded under the auspices of UNESCO in 1960, with national societies as members. Interestingly, one of the earliest contributions of IFIP was defining the ALGOL 60 programming language, which proved to be, in turn, the foundation for several other imperative languages such as C and Pascal. CSI represents India in the IFIP General Assembly, and the Immediate Past President, Mr. M D Agrawal will represent CSI in the forthcoming meeting of the GA. SEARCC South East Asian Regional Computer Confederation (SEARCC) is a regional association of the national societies of Asia Pacific, founded in 1978 at an IFIP meeting at Singapore. The CSI President is also the President of SEARCC for 2012, and I will be representing CSI in the activities of SEARCC. At this time, a priority for SEARCC is to increase its membership as it is limited to about 6 countries. SEARCC also is seeking to organize funds for its operations. IEEE Computer Society and ACM CSI has been negotiating with IEEE and IEEE Computer Society for the renewal of existing MoUs with them, and with ACM for a new MoU. While the process of finalization of the MoU with ACM is continuing, the June ExeCom approved the signing of the MoU with IEEE Computer Society. This MoU, the details of which will be available on the CSI Website as soon as it is formally signed within about a month, has a number of innovative features that will benefit CSI members, especially students and researchers. We hope to start its implementation from August BASIS From 2009, CSI has been a member of Business Action to Support the Information Society (BASIS), which is a body of the International Chamber of Commerce, set up after the 2003/2005 World Summit on the Information Society (WSIS) at Geneva/Tunis. BASIS contributes significantly to Internet Governance through its presence at the Internet Governance Forum (IGF) and other similar UN and international processes. BASIS offers opportunities for CSI to articulate the civil society perspective at the IGF. In the June ExeCom, CSI decided to renew its membership in BASIS. Theme: Digital Image Processing This month's theme for CSIC is Digital Image Processing. Image Processing has been one of the relatively recent computational fields, emerging towards the last decade of the twentieth century. It has grown rapidly on account of the explosive growth in its applications. In essence, Digital Image Processing is an instance of digital signal processing, the signal in this case being any image source such as a video frame, photograph, photomicrograph, or satellite imagery. Today, Digital Image Processing is employed in a wide variety of applications. In medicine, image processing is extensively used in diagnostics and pathology, and is an integral component of such popular techniques as radiology, nuclear medicine, endoscopy, and microscopy. In Remote Sensing, image processing provides an extensive set of algorithms for the analysis of satellite imagery, which could be multi/super/hyperspectral for military applications as well as for precision agriculture, natural resource management, and coastal zone management. In our day-to-day lives, Digital Image Processing has been applied to solve interesting problems. Face recognition and biometrics is popularly available, including as a built-in option in some laptops and tablet computers. In transportation security, face recognition is now capable of seeing through disguises and can isolate the underlying physiognomy of individuals. Advanced scanners are able to provide three-dimensional images of individuals in some airports for security scanning. Video-based VOIP conferencing is a popular feature on Smart Devices such as mobiles and pad computers. Automatic number plate recognition is being used in many parts of the world, including India, to identify vehicles involved in traffic violations. Driverless cars, now being piloted by several companies employ advanced forms of real-time image processing. Among all its applications, it is perhaps military applications that are most advanced in the digital computing domain. Many fighter cockpits employ Head-Up Displays which project all flight parameters on to the normal outside view surface instead of pilots having to look downwards at their flight instruments. Another important use of digital image processing is in Unmanned Aerial Vehicles (UAVs) which are formidable weapons in today's wars. Since the command-and-control centers for UAVs are remotely operated, the entire video captured by the UAV has to be processed and transmitted in real time. UAVs are also used in peacetime applications, for instance in surveillance after natural disasters. Given the nature of the terrain, lack of line-of-sight tracks, and limited satellite bandwidth, video transmission in UAVs require the use of highly efficient algorithms. Finally, the entertainment industry is another major user of image and video processing technologies. Most feature films are now shot and edited using digital technologies. 3-D modeling and animation technologies have been used side-by-side with image processing technologies for films. Commercial broadcasting and cable TV also make extensive use of digital image processing, especially with 3-D television. Given one or two cameras in most mobile phones, and a million pictures being uploaded daily to some of the social networking sites, it is clear that digital storage, processing, distribution, and use of images will continue to increase. Digital Image Processing, once the domain of a few select research applications, is now a ubiquitous technology, touching everyone's lives. With greetings Satish Babu President CSI Communications July

6 Editorial Rajendra M Sonar, Achuthsankar S Nair, Debasish Jana and Jayshree Dhere Editors Dear Fellow CSI Members, We are happy to release special issue on image processing covering articles from distinguished experts in the arena. When we talk about image processing, we usually mean digital image processing, but analog and optical image processing also emerge. Image can be enhanced in quality with increased contrast, compressed in size with minimum deterioration, restored with reduced blurring and also can be used for extraction of useful features and characteristics so that machines can visualize. Computer vision, biomedical imaging, pattern recognition, astronomical and geospatial imaging, content based image search are all in the platter. Image processing algorithms apply point, local, global operations on a digital image like detection of edges, elimination of high frequency noise, contrast stretching. Image can be enhanced in quality with increased contrast, compressed in size with minimum deterioration, restored with reduced blurring and also can be used for extraction of useful features and characteristics so that machines can visualize The issue starts with Cover Story titled An Algebraic Method for Super Resolution Image Reconstruction by Prof. Bhabatosh Chanda of Indian Statistical Institute, Kolkata. In his article, Prof. Chanda presents a technique of image reconstruction of super resolution images as an unconstrained optimization problem. Technical Trends section is enriched with an article on Applications of Image Processing in Industries by Dr. Tanushyam Chattopadhyay, Brojeshwar Bhowmick and Aniruddha Sinha of TCS Innovation Labs, Kolkata. In Research Front section, Dr. Saurav Basu of Carnegie Mellon University has articulated his thoughts on Some Upcoming Challenges in Bioimage Informatics on the backdrop of the conceptual ecosystem of a dedicated image analysis/computer vision framework. R P Ramkumar of Mahendra Institute of Technology and Dr. S Arumugam of Nandha Educational Institutions have presented biometric identification through iris in their article on Accurate Pupil and Iris Localization using Reverse Function. In Article section, Hema Ramachandran of College of Engineering, Trivandrum has presented write-up on Image and Video Processing Toolbox in Scilab an effective alternative to Matlab. In another article titled Importance of Shifting Focus in Solving Problems, Dr. Pramod Koparkar has shown interesting application of image synthesis technique in geometric modelling. Another article authored by Hareesh N Nampoothiri of University of Kerala has covered Distance Units in CSS3, new standard for cascading style sheets recommended by W3C. Practitioner Workbench column has a section titled Programming.Tips() and it provides an interesting write-up on Multidimensional Plots in Matlab for Data Analysis by Baisa L Gunjal of Amrutvahini College of Engineering Sangamner and Dr. Suresh N Mali of Singhgad Institute of Technology and Science, Maharastra. The other section called Programming.Learn("Python") under Practitioner Workbench covers guidelines on how to read from text file and how to write results to a file using Python. Challenges in Bioimage Informatics on the backdrop of the conceptual ecosystem of a dedicated image analysis/computer vision framework. Information Security section of the Security Corner feature has an interesting article on Security and Privacy, which throws light on privacy concerns in the age of social networking. Another section called IT Act 2000 under Security Corner comes with a write-up by Mr. Subramaniam Vutha, wherein he explains the concepts of offer and its acceptance in the context of online shopping through a dialogue between a legal expert and an IT executive. This time we are dropping CIO Perspective and HR column but next time you will surely have them. As usual there are other regular features such as Brain Teaser, Ask an Expert, ICT@Society and Happenings@ICT. CSI Reports and CSI News section provide event details of various regions, SIGs, chapters and student branches. Please note that we welcome your feedback, contributions and suggestions at csic@csi-india.org. With warm regards, Rajendra M Sonar, Achuthsankar S Nair, Debasish Jana, and Jayshree Dhere Editors CSI Communications July

7 Cover Story Bhabatosh Chanda Electronics and Communication Sciences Unit, Indian Statistical Institute, Kolkata An Algebraic Method for Super Resolution Image Reconstruction Abstract: Super resolution (SR) image reconstruction is a critical problem arising due to hardware limitations. It is well known that the SR problem is ill-posed and inverse methods are not directly applicable. In this article, we present the SR image reconstruction method as an unconstrained optimization problem and solve it using iterative technique with proper choice of regularization term. Experimental results show the efficacy of the method. Keywords: Super resolution imaging, illposed problem, least-square estimation We present the SR image reconstruction method as an unconstrained optimization problem and solve it using iterative technique with proper choice of regularization term. Introduction Whenever we capture images with a camera our intention is almost always to have high-resolution (HR) images as these are good for both viewing as well as computer processing. However, to have the desired HR image is not always possible and the major bottleneck is the hardware limitations. A direct method for increasing pixel resolution would be by dense sensor manufacturing technique. However, increasing chip size or packing more sensors in a single chip leads to increase in capacitance, which makes it difficult to speed up the charge transfer rate. Thus, using image restoration technique to obtain a HR image from the observed low-resolution (LR) image(s) has become a promising solution to the problem. This is simply a method of increasing pixel resolution synthetically [1,5] and is called super resolution image reconstruction or super resolution imaging. The topic remains an area of active research because of its many interesting applications including synthetic zooming of region of interest in forensic, surveillance, remote sensing, and conversion from NTSC video signal to HDTV signal. The SR methods can be broadly categorized into two groups: (i) generating HR image from multiple LR frames known as multi-frame SR, and (ii) generating HR image from a single LR image known as single-frame SR. Here we will focus only on multi-frame SR using an algebraic technique. Generating LR Image Before developing the algorithm for generating HR image from LR images, let us mention again that the SR imaging means the enhancement of the spatial resolution only and not of gray level or color. Moreover, the enhancement in image quality by SR methods is possible if and only if LR images were sampled at a rate lower than the Nyquist rate. Second, the multi-frame SR methodology Fig. SR-1: An example showing subsampling and fusion of subsamples to obtain super resolution version. (a) Original analog signal, (b) sampled version of the signal in (a) with a relative shift between (i) and (ii), (c) train of samples obtained from b(i) and b(ii) respectively, and (d) train of samples by combining c(i) and c(ii) resulting in super resolved signal compared to (c). CSI Communications July

8 f(x) f i f Down sampling factor = 2 g decimation matrix that describes the downsampling process such as: x i x i x i - Δx/2 + Δx/2 (a) works only when we have sufficient number of LR images of the same scene that are different from one another because of motion or blurring or any other physical reason. Let us first consider an example of a one-dimensional signal as shown in Fig. SR-1(a). Two different sampling instances of this signal at the same sampling rate but with a little offset are shown in Figs. SR-1(b)i and SR-1(b)ii. Thus, we get two LRsampled signals [Figs. SR-1(c)i and SR-1(c) ii]. Given these two signals, the objective of SR technique is to reconstruct the HRsampled signal as shown in Fig. SR-1(d). It is obvious that the reconstruction of HR signal from the LR signals needs mutual registration of the latter ones. For mutual registration of the sampled signals [Figs. SR- 1(c)i and SR-1(c)ii] they have to be mutually correlated, which is not apparent if (shifted) delta functions are used for sampling as described in the figure. However, according to digital image acquisition technology, a sample value is not really an instantaneous value of the signal, but approximately the average intensity over the sampling interval as revealed in Fig. SR-2(a), where the sample value f i may be defined as: x i +Δx/ 2 f i = 1 f ( x ) d x Δx x i Δx/ 2 where x is the sampling interval. If we consider the samples of Fig. SR-1(d) as the HR version of the signal and denote them by f i (i = 0, 1, 2,...), then the LR samples [see Fig. SR-1(c)] may be obtained by down sampling f i as g 1 j = 1 2 (f +f ) and 2j 2j+1 g j 2 = 1 2 (f 2j+1 +f 2j+2 ) assuming that the x Fig. SR-2: (a) Value of sample corresponds to the area under the curve. (b) Subsampling in two-dimension by a factor of 2. resolution of g j is half of that of f i and the relative shift is one sample. Similarly, in two-dimension, one of the LR images g(r,c) may be obtained by averaging the corresponding 2x2 pixels of HR image, i.e. g(r,c) = 1 (f(2r,2c) +f(2r,2c + 1)+f(2r + 1,2c)+f(2r + 1,2c + 1) 4 Whenever we capture images with a camera our intention is almost always to have high resolution images as these are good for both viewing as well as computer processing. (b) This is illustrated in Fig. SR-2(b). Thus, g(r,c) is the downsampled image of f(r,c), where downsampling factor is 2. With this understanding let us state the super resolution imaging problem as follows. Problem Definition Suppose a two-dimensional array of M X N CCD sensors cover the entire LR image plane resulting in observed images of M rows and N columns. Corresponding HR image is of size sm X sn, where s is the downsampling factor. A column vector g of size MN is generated from the given LR image matrix {g(r,c)} by raster scan of the matrix. A column vector f of size s 2 MN corresponding to the HR image may be obtained in a similar way. Suppose we have n such LR images g i (i = 1, 2, 3,..., n), and then assuming that each LR image is corrupted by signal independent additive noise we can represent the i-th observed LR image as: g i = SB i V i F + η i for i = 1, 2, 3,..., n (sr-1) where V i is a transformation matrix of size s 2 MN X s 2 MN representing motion of the objects observed in i-th LR image, B i is a s 2 MN X s 2 MN matrix representing blurring including defocusing and atmospheric effect, and finally S is the MN X s 2 MN s(r+1) 1 s(c+1) 1 g(r,c) = 1 f(x,y) (sr-2) s 2 x=sr y =sc An example of conversion of a HR image to corresponding LR image by down sampling by a factor of 2 is shown in Fig. SR-2(b). To understand the process let us consider these transformation matrices separately. The motion matrix V i includes mainly global translation and rotation. An example of translational motion is shown in Fig. SR-3, which results in shifted sampling as shown in Figs. SR-1(b)i and SR-1(b)ii. The motion parameters are usually estimated by choosing one of the LR frames (arbitrarily) as a reference frame and the shift (motion vector) of the other LR frames are computed through registration and interpolation with respect to the reference frame. Blurring matrix B i represents the degradation process (e.g. defocusing, motion blur etc.). It may also include the averaging part of the decimation matrix S [Equation (sr-2)]. In that case, the only task of S is to pick up appropriate pixels from shifted blurred HR image. Thus B i and S are same for all LR images and only V i varies from one LR image to another. So we can rewrite Equation (sr-1) without loss of generality, similar to image restoration [2], as g i = H i f + η i for i = 1, 2, 3,..., n (sr-3) where H i = SBV i of size MN X s 2 MN. Hence, the problem here is to estimate a HR image f given a set of (precisely n number of) LR...using image restoration technique to obtain a high-resolution (HR) image from the observed low-resolution (LR) image(s) has become a promising solution... images g i, the matrices H i s, and knowledge about the noise term η i. As n < s 2, i.e. the total number of available LR image pixels is less than that of unknown HR image pixels, the super resolution image Original HR grid Shifted HR grid (due to motion) Fig. SR-3: Shows the relatively shifted sampling grid of HR image. Black and red lines represent reference and shifted grid lines respectively. CSI Communications July

9 ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( (a) (b) (c) Fig. SR-4: Result of SR reconstruction by factor 5. (a) One of the LR images, (b) zooming by bilinear interpolation, and (c) result of a well-known SR imaging [4]. reconstruction, in general, is an ill-posed problem. That means Equation (sr-3) has infinitely many solutions. Second, since size of H i is extremely large, direct inverse methods are computationally infeasible. There are many different approaches to solve this problem. Here we discuss an algebraic method, namely constrained leastsquare estimation. Constrained Least-Square Estimation Suppose we have the reconstructed HR image f. This would be a good estimate if the difference between the observed LR images and the redegraded downsampled estimated image f be zero. Because of the ill-conditioned nature of the solution, a number of f may be available that would satisfy this criterion. From among these f, we need to select a particular one. For this purpose we may employ a selection criterion, which can obtain a particular f that satisfy some criterion of goodness or quality measure of image Qf 2 subject to the constraint that total residual norm between the LR images and the redegraded down sampled estimated image be minimum. In other words, we try to obtain a solution f as..according to digital image acquisition technology, a sample value is not really an instantaneous value of the signal, but approximately the average intensity over the sampling interval as reveled.. minimize n i= Qf 2 such that g H ^ f 2 η 2 < ε i i i The quality measure operator Q could be first or second order derivative operator. In that case, quality measure Qf 2 emphasizes on the smoothness in the HR image, and n the data error term g i H ^ 2 2 i f η i i= 1 should be less than a predefined tolerable error. Since obtaining solution to a constraint minimization problem is a nontrivial task, we reformulate the above problem as an 150 unconstrained minimization problem as n J( ) = Q ^ f ^ f +λ( g i H ^ i f ηi ) i= 1 (sr-4) where λ is the Lagrange multiplier, commonly known as the regularization parameter in the super-resolution literature, which maintains a trade-off between the regularization term Qf 2 and the error in estimation. J(f ^ ) of equation (sr-4) is a quadratic and convex function of the estimated image f, so the f for which J(f ^ ) is minimum may be obtained by differentiating J(f ^ ) and equating the result to zero. Hence, we can write n n ( H T H i +γq T Q) ^ f = H T g i (sr-5) i= 1 i= 1 that leads to an iterative solution for f as [3] n k f +1 = f k +α H T i g i H ^ k i f ) γq T Q^ f k ] (sr-6) i= 1 where k is the number of iteration and α controls the convergence. An experimental result of SR method, described here, is shown in Fig. SR-4 [4]. References [1] Chaudhuri, S, ed. (2001). Super- Resolution Imaging, Kluwer, Norwell, MA. [2] Chanda, B and DuttaMajumder, D (2011). Digital Image Processing and Analysis, 2nd edition. PHI Learning, New Delhi. [3] Park, S C, et al. (2003). Superresolution image reconstruction: A technical overview. IEEE Signal Processing Magazine, 5, [4] Purkait, P and Chanda, B (2011). Morphologic gain-controlled regularization for edge-preserving super-resolution image reconstruction, Signal, Image and Video Processing. Springer, London. [5] Rajan, D, et al. (2003). Multiobjective super-resolution: Concepts and examples. IEEE Signal Processing Magazine, 5, n ] ) About the Author Dr. Bhabatosh Chanda is a Professor in the Electronics and Communication Sciences Unit at Indian Statistical Institute, Kolkata. He received BE in Electronics and Telecommunication Engineering and PhD in Electrical Engineering from University of Calcutta in 1979 and 1988 respectively. His research interests include Digital Image Processing, Pattern Recognition, Computer Vision, Image Analysis, Mathematical Morphology, and AI techniques in Computer Vision. He has been actively working in the field for more than 25 years. He has authored more than 125 technical articles in refereed journals and conferences. He has written a book titled Digital Image Processing and Analysis published by PHI Learning and eight edited volumes including conference proceedings. He has received Young Scientist Medal from Indian National Science Academy in 1989, Computer Engineering Division Medal from the Institution of Engineers (India) in 1998, and Vikram Sarabhai Research Award from Physical Research Lab in He is also a recipient of UNDP fellowship ( ) and Diamond Jubilee fellowship of National Academy of Science (1992). He is Fellow of Institute of Electronics and Telecommunication Engineers (IETE); the National Academy of Science, India; Indian National Academy of Engineering; and International Association of Pattern Recognition. CSI Communications July

10 Technical Trends Dr. Tanushyam Chattopadhyay, Brojeshwar Bhowmick, and Aniruddha Sinha Innovation Lab, Tata Consultancy Services, Kolkata Applications of Image Processing in Industries Image processing (IP) is used in diversified application areas in the industries. In this article, we shall describe three different perspectives of industrial applications of image processing: computer vision, video technologies, and 3-D reconstruction. We shall also describe some industry-specific IP applications based on our experiences as a part of service industry. Computer Vision Computer vision-based applications are used in several manufacturing processes like manufacturing of delicate electronics components [21], quality textile production [1], metal product finishing [27], glass manufacturing [15], machine parts [10], printing products [26] and granite quality inspection [23], integrated circuits manufacturing [12] and many others. A generic diagram of computer vision system is depicted in Fig. 1. Any computer vision related applications require the following major components: Image acquisition, Image processing, Feature extraction, and Decision making. Part Sensor Lighting Sensor Frame Grabber Industrial PC Inspection Software Digital IO Fig. 1: Generic computer vision system in industries Image acquisition A scene in the external world is illuminated by light source. Image is formed on the receiver or the camera. The radiance reflected by an object depends in the spectrum of light source and surface properties of the object. The geometry of the image formation depends on the structural composition of the camera. Light source causes electromagnetic radiation. The spectrum of the light ranges from radio waves to X-ray and gamma waves including the visible spectrum [29]. Depending on the field of application, various ranges of artificial and natural light sources are used. Some of these are summarized below: (i) Infra-red (IR) - The IR radiation is mainly of three types, viz. far IR, thermal IR, and near IR. Applications of thermal IR include remote sensing, night vision without any help of illumination, heat sensing etc. IR and near IR light sources are used to project invisible illumination on objects, and then capture the images using specialized CMOS sensors capable of being exited in the range of 10 micrometer and 1 micrometer. These are used in medical and automotive applications where the IR spectrum aids in image processing with controlled IR light source. (ii) Visible - Apart from the natural light source, the visible spectrum is generated by many types of artificial sources which include incandescent lamps, metal vapor lamps, xenon lamps, fluorescent lamps, light emitting diodes (LED), and lasers [3]. Most of the industrial applications in mediaentertainment and surveillance use the visible spectrum of light source. Microscopic and Satellite image processing are two specialized domains which use multispectral analysis. In case of microscopes, the object dimensions are in microns and the images are captured after enough magnification so that it is visible by human eye. On the other hand, for satellite image acquisition, the objects are on the earth surface and the images are taken from tens of kilometers away using specialized cameras with resolutions ranging from 1 meter to few 100 meters. Surface properties and the texture affect the radiant light entering the camera from the object surface. Diffused and specular reflections from object surfaces provide major challenges in image processing especially in object segmentation. There can be multiple reflections, secondary illuminations, and shadow formations causing the image content analysis challenging. Camera properties including lens structure and camera sensor of the camera affect image. Inaccuracy in the lens structure or sensors can cause image distortions. Radial distortion includes barrel and pincushion distortion. Chromatic aberration may create color fringe effect in the periphery of the images. Image processing techniques [23] are employed to correct these distortions. The most popular sensors of the camera are Complementarymetal-oxide-semiconductor (CMOS) and Charge-coupled-device (CCD) [3]. Image distortions can be corrected using camera internal parameters [25]. Image processing Many a times, the captured images need to be processed first before using into a computer vision system. Some of such problems are: (i) Blurring, (ii) perspective transformations, and (iii) noise. Some examples are shown in the Fig. 2. These images are sometimes required to be segmented so that the region of interest (RoI) can be identified. Fig. 2: A few sample images which require preprocessing prior to process. (a) Perspective transformation and (b) Reflection Feature extraction It is preferable to select nonoverlapping or uncorrelated features or characteristics of an image [16], so that better classification can be achieved. Examples of such features include size, shape, pose, texture, and color. Once the features are computed, they are usually used as a machine learning module to learn the system. The machine learning can be either supervised or unsupervised. Sometimes, the feature set is optimized by reducing the feature dimension using PCA or other similar methods. Decision making Once the features are obtained they are used to reach a decision. In some applications, decision is taken by the machine itself and sometimes the method aid humans to take any decision. Video Technologies Video technology is mostly used in media and entertainment industries. Surveillance video feed analysis is also an important area in security purpose. Some major steps for deploying any video-based system are: Video capturing: Video is captured using almost similar manner as how an image is captured. Human eye cannot perceive changes if any camera captures more than 15 image frames per second. Video Compression: Typically, any video of even 720x576 resolution captured at a 25 FPS rate requires almost 15 MBps to communicate over any channel, thereby CSI Communications July

11 needing video compression. Recent technical trends show a requirement to deploy such video encoder-decoder on an embedded platform. The recent video encoders are capable of compressing the video at a low-bit rate even keeping the video quality significantly good but at the cost of computational complexity. So it is a challenge in the industry to deploy them on an embedded platform. Video Transcoder: Recent Consumer Electronics trends show the requirement of rendering same video content over different display devices like ipod, iphone, ipad, High Resolution TV, and smartphones. Different devices support different CODEC and different resolutions. It is required to change the format (in terms of resolution, aspect ratio, and video coding standard) from one to another (transcoding). 3-D Reconstruction Three-dimensional (3-D) reconstruction from multiple two-dimensional (2-D) images or images captured using structured lights and depth sensor has various applications in the field of manufacturing (dimension measurements), tourism (virtual 3-D walk through), gaming (virtual reality and augmented reality), media and entertainment (3-D movies), medical imaging (Computerized Tomography and Magnetic Resonance Imaging) and many other areas. Various types of 3-D reconstruction are given below: Multiple 2-D images 3-D reconstruction from multiple images is an active area of research where the main objective is to obtain threedimensional Euclidean structure of a desired scene from general photographs. Projective geometry [8] is the basic mathematical framework that is used to establish different relations among multiviews which essentially give rise to the 3-D structure. Image formation from the 3-D scene is a nonlinear process. Multiview reconstruction employs epipolar geometry [8] to have relations between two images, which is further carried for projective reconstruction. The Euclidean structures can be obtained by providing additional information about the scene into projective reconstruction or through a nonlinear optimization called Bundle Adjustment. This reconstruction normally happens in sparse, which means a subset of the scene points is reconstructed. Fig. 3 shows the images that are part of a pillar of Qutub Minar and its sparse reconstruction along with camera estimation. Fig 3: Images and sparse reconstruction of pillar of Qutub Minar Given this the geometry recovery for sparse points enable us to figure out how the structure looks like at coarse level, what is the relation of position between cameras and structures and the camera calibrations. All these information are carried forward to a more refined reconstruction, dense reconstruction using Space Carving [11] or Volumetric Graph cuts [28]. These reconstruction techniques deal more with finer details of the structures as shown in Fig. 4. Fig. 4: Dense reconstruction using Space Carving Structured lights and depth sensor Kinect, a motion sensing input device by Microsoft, has redefined the Human Computer Interaction (HCI). Among the most popular applications using Kinect are the X-box games as well as gesture-based interactions. The skeleton movement using Kinect also aids in recognizing gaits and action simulations in many applications. Similar to 3-D point cloud, with the help of Structure light, a predefined light pattern is thrown to the object and camera captures the scene. Fusion of depth sensor and 2-D image Kinect point clouds can directly be used to have surface models, which Kinect- Fusion [19] does. There is some scope of the high-resolution point cloud generation from these point clouds which is of resolution 640x480. The resolution could be enhanced by incorporation of two or more HD cameras with Kinect as suggested in [18]. Epipolar geometry is used to connect the Kinect with HD cameras and Graph-Cut based Energy Minimization is used to get the geometrically correct high-resolution point cloud as shown in Fig. 5. The high-resolution point cloud has 4x more vertices than the original point cloud. This could be used as more accurate modeling of an object where the Kinect resolution is not sufficient. Fig. 5: High-resolution cloud point using normal HD images and Kinect cloud points Applications We have classified the applications of image processing based on the area of client s business focus that drives the requirements. Automotive In recent days, the cars are getting connected to Internet and amongst themselves. This wireless connectivity has opened up a wide range of infotainment applications and transformed the driver assistance [13] to a completely new dimension. A summary of some of the applications are given here. (a) Driver safety monitoring - The most unobtrusive method uses IR camera mounted in front of the driver near the windshield and monitors the driver s head movement and closure of eyes for possible sleep detection [7]. This is used as an assistive measure to alert the driver in case of inattentiveness. (b) Driver assistive solutions - Camera and radar-based sensors are quite popular for assisting the driver to detect traffic signs and pedestrians, indicate lane departure, viewing blind spots, parking etc. [17] The standard Vienna Convention compliant signs are detected by the Traffic-sign recognition (TSR) system provided by Mobileye [14]. (c) Recently various smartphone applications have been developed to detect the lane markers and assist in navigation by overlaying graphical directions on top of the road maps. It is a real challenge in developing countries like India to provide driver-assistive solutions in the environment where the road signs are not standardized and the lane markers are not available in most of the places. Few prototypes have CSI Communications July

12 been attempted in India on pedestrian detection using stereo camera mounted on a car and measuring the distance of the same [24]. This has been tested at 20 kmph in normal daylight detecting people at a distance within 100 feet. (d) Display cluster test automation - With the advancement of infotainment solution, the car dashboard displays are converted to electronic displays as shown in Fig. 6. In order to automate the testing for these displays, image processing techniques [9] are applied to compare the real-time video captured in camera against the prestored templates. Fig. 6: Sample electronic display in car dash-board Retail Retail stores are nowadays getting automated to allow smart check out. But most of the existing methods of those retail stores use RFID and other sensor-based technologies. But some interesting image processing applications are used in garment sections of such retail stores. Some such examples are described below: (i) Garment color detection: Sometimes users search for a garment of a particular color. The customer selects the shirt from a rack of shirts of different size. Now the sales person needs to search whether there is any shirt of that color and that size in the inventory. The sizes are indexed based on some numbers. Image processing methods helps to define the major colors of a shirt and index it into database accordingly. (ii) Magic Mirror: Customers appreciate to see how they look with their garment. But sometimes it is unhygienic to wear the garments to see how they look. So image processing-based approaches can provide an alternative where the customer just holds the garment in front of him and the magic mirror (a PC indeed) shows him how he looks after wearing this garment. A typical use case is that a customer would upload a image of his/hers and there would be a system which will identify various pattern (on face) in that image, viz. eyes, lips etc. The customer can try various kinds of lip color on that image and make a better informed purchase decision. The main problem of such a system is that the images are of different types and often due to mismatch in identification of the face, the lipstick for instance gets applied on the chin or eyes. Insurance Health and Car Insurance companies are nowadays keen to use image processing. Some of the image processing applications are described here. Health Insurance companies usually store the scanned copies of the medical prescription. But the person doing this scanning job sometimes skews the document while scanning it. That significantly affects the compression performance of such documents. Moreover, these prescriptions are required just to keep the texts in the database. So a proper binarization method is required to get a good compression. In Car Insurance, the Vehicle Identification Numbers (VIN) is imposed to be used by the National Highway and Traffic Safety Administration of the USA. So, in case of any damage or accident, the VIN of the concerned vehicle is needed to be sent to the insurance company to process any insurance claim. On the other hand, with the advancement of the consumer electronics technology most of the mobile handsets are now equipped with a digital camera. Smartphones are also capable of doing some processing on the embedded hardware platform of the mobile handset like Android. So the insurance companies are thinking of providing some application for the smart phone users that can recognize the characters from the VIN images captured by the common users. These 17 alpha-numeric characters cannot be sent by simply typing them because of authentication issues. Some research on such problems can be found from [5]. Medical Image processing is commonly used to process the MRI, CT scan images for quite a long time. Nowadays, surgical instrument manufacturing companies also use image processing in different ways. (i) Some surgical instrument companies usually rent their instruments to some hospitals and after the operation is done the hospitals return it to the manufacturers. But the hospitals frequently misplace the instruments or sometimes forget to return them. Once the instruments are returned they need to sterilize them. So an image processing based method was deployed there to monitor whether all the instruments are placed properly or not. Details of such system can be found from [2]. (ii) Some industries, on the other hand, want to see whether the doctors are using the instruments properly in the operation theater or not. So there is a video search method that takes an instrument image as input and returns the time stamps where the instrument has been used in the video. Details of such a system can be found from [4]. Media and entertainment As the TV is connected with the Internet, there is a need to extract the contextual information from any TV video and then fetch related information from the web to provide a true connected TV experience to the viewers. Number of applications can be developed using those information to improve the user experience [6]. Though Digital TV is popular all over the developed countries, even today in India more than 90% of TV households have analog broadcast cable TV. Thus, unlike digital TV transmission it is not possible to automatically get contextual information from any metadata. In the proposed method, we have used two context information from the TV video, viz. (a) the channel a user is viewing and (b) the synthetic text in the video like subtitle of a movie or scorecard of a sports show or news ticker of a breaking news in a news channel. The current viewing channel is recognized by matching the channel logo against the set of preexisting channel logo templates. The text in a TV channel is extracted by text region identification followed by preprocessing of the text regions and performing Optical Character Recognition (OCR) on the text regions. Next generation mobile technologies In recent days, mobile phones are equipped with high-resolution cameras which range up to 8-10 Megapixels. Among many image processing applications, recently these mobile phones are being experimented for monitoring various physiological parameters in human beings. It has been reported that by analyzing variation of color in the images captured from the finger tips using a normal mobile phone camera, one can measure blood oxygen saturation, breathing rate, and cardiac R-R intervals [22]. Remote education Distance education is gaining popularity in rural areas of developing countries due CSI Communications July

13 to the shortage of teachers. In order to support a variety of curriculum for different states in a diverse country like India along with various languages, there is a need for intelligent technologies to create a deployable remote education solution. The main challenge in the distance education solution is the dissemination of content and teacher-student interaction. Internet protocol based solutions are yet to pick up in India due to the lack of infrastructure and unavailability of high-bandwidth network connection. Solutions based on existing TV broadcast network is proposed where video multiplexing is done based on video content in order to support a large number of consumers using a few TV channels [20]. References [1] Bahlmann, C, et al. (1999). Artificial Neural Networks for Automated Quality Control of Textile Seams. Pattern Recognition, 32, [2] Chaki, A and Chattopadhyay, T (2009). An Automatic decision support system for medical instrument suppliers using fuzzy multifactor based approach. The fifth annual IEEE Conference on Automation Science and Engineering (IEEE CASE 2009), 8, [3] Chaki, A, et al. (2010). A Comprehensive Market Analysis on Camera and Illumination Sensors for Image Processing and Machine Vision Applications. International Conference on Computational Intelligence and Communication Networks (CICN), 11, [4] Chattopadhyay, T, et al. (2008). An Application for Retrieval of Frames from a Laparoscopic Surgical Video Based on Image of Query Instrument. TENCON 2008, IEEE Region 10 Conference, 11, 1-5 [5] Chattopadhyay, T, et al. (2012). On the Enhancement and Binarization of Mobile Captured Vehicle Identification Number for an Embedded Solution. 10th IAPR International Workshop of Document Analysis, Australia. [6] Chattopadhyay, T, et al. (2012). Value Added Services for Connected TV. LAP LAMBERT Academic Publishing, ISBN-13: [7] Chidanand Kumar, K S and Bhowmick, B (2009). An Application for Driver Drowsiness Identification based on Pupil Detection using IR Camera. International Conference on Human Computer Interaction, IIIT Allahabad. [8] Hartley, R I and Zisserman, A (2004). Multiple View Geometry in Computer Vision, second edition, CUP, Cambridge. [9] Huang, Y, et al. (2009). Model-based testing of a vehicle instrument cluster for design validation using machine vision., Measurement Science and Technology, 20(6). [10] Ker, J and Kengskool, K (1990). An Efficient Method for Inspecting Machine Parts by a Fixtureless Machine Vision System. Vision '90 Conference. [11] Kutulakos, K N and Seitz, S M (2000). A Theory of Shape by Space Carving. International Journal of Computer Vision, 38(3), [12] Li, H and Lin, J C (1994). Using Fuzzy Logic to Detect Dimple Defects of Polisted Wafer Surfaces. IEEE Transactions on Industry Applications, 30, [13] Masikos, M, et al. (2011). EcoGem - Cooperative Advanced Driver Assistance System for Green Cars. Advanced Microsystems for Automotive Applications. [14] Mobileye Technologies Ltd. (Nicosia, Cy), Bundling Night Vision And Other Driver Assistance Systems (Das) Using Near Infra Red (Nir) Illumination And A Rolling Shutter. United States Patent Application [15] Novini, A R (1990). Fundamentals of Machine Vision Inspection in Metal Container Glass Manufacturing. Vision '90 Conference. [16] Oyeleye, O and Lehtihet, E A (1998). A Classification Algorithm and Optimal Feature Selection Methodology for Automated Solder Joint Inspection. Journal of Manufacturing Systems, 17, [17] Parnell, K (2003). Driver assistance systems - real time processing solutions. Intelligent Vehicles Symposium Proceedings IEEE, 6, [18] Patra, S, et al. (2012). High Resolution Point Cloud Generation from Kinect and HD Cameras using Graph Cut. VISAPP, 2, [19] Richard, A, et al. (2011). KinectFusion: Real- Time Dense Surface Mapping and Tracking. IEEE ISMAR, IEEE, 10. [20] Saha, A, et al. (2012). Embedding Metadata in Analog Video Frame for Distance Education. International Journal of e-education, e-business, e-management and e-learning, 2(1), 11. [21] Sanz, J L C and Petkovic, D (1988). Machine Vision Algorithm for Automated Inspection of Thin-Film Disk Heads. IEEE Trans. on PAMI, 10, [22] Scully, C G ;(2012). Physiological Parameter Monitoring from Optical Recordings With a Mobile Phone. IEEE Transactions on Biomedical Engineering, 59(2), [23] Shafarenko, L, et al. (1997). Automatic Watershed Segmentation of Randomly Textured Color Images. IEEE Trans. on Image Processing, 6, [24] Sinharay, A, et al. (2011). A Kalman Filter Based Approach to De-noise the Stereo Vision Based Pedestrian Position Estimation. UKSim 13th International Conference on Modelling and Simulation, [25] Szeliski, R (2010). Computer Vision: Algorithms and Applications. Springer publication, New York. [26] Torres, T, et al. (1998). Automated Real-Time Visual Inspection System for High-Resolution Superimposed Printings. Image and Vision Computing, 16, [27] Tucker, J W (1989). Inside Beverage Can Inspection: An Application from Start to Finish. Proc. of the Vision '89 Conference. [28] Vogiatzis, G, et al. (2007). Multiview Stereo via Volumetric Graph-Cuts and Occlusion Robust Photo-Consistency. IEEE Trans. Pattern Anal. Mach. Intell., 29(12), [29] spectrum n About the Authors Dr. Tanushyam Chattopadhyay got his PhD degree from Jadavpur University. He has received the BSc in Physics from Visva Bharati and completed his MCA from Bengal Engineering College, Shibpur, India, in 1998 and 2002, respectively. He has started his career as research personnel in Indian Statistical Institute, Kolkata, and currently he is working as a Scientist in Innovation Lab, Kolkata, TCS. His areas of interest include Image and Video analytics, Video compression, Video Security, Video Retrieval, Video summarization, Image and video Pattern Recognition, Image, Speech, processing. He has nearly 40 papers in peer reviewed international conference and journals in this field. He is author of a book on Value Added Services for connected TV and also some book chapters. He has received many awards like University Medal in masters, CSI YITP special mention award at the national level after being the winner at the regional level, WWW best software, TCS Top 10 coder, TCS patent champion, TCS young Innovator etc. Brojeshwar Bhowmick is a PhD Scholar at IIT Delhi and scientist at innovation lab, TCS. He has 7 years of experience both in academic and industry, like, Indian Statistical Institute, IIT Delhi, Avisere Technology(now Videonetics), and Innovation lab TCS. His research areas are Computer Vision, Machine Learning, Image and Video Understanding, Multi-view Geometry. Currently he is doing research in Geometric Vision, 3D Reconstruction and Machine Learning at IIT Delhi. He has more than 15 papers in peer reviewed international conference and journals in this field. He wrote a Chapter in a Digital Image Processing book. He also awarded Young IT Professional Special Mention Award in CSI Kolkata. Aniruddha Sinha B.E. in Electronics and Telecommunication Engineering from Jadavpur University, Kolkata, India, in June-1996 and M.Tech in Electrical & Electronics Communication Engineering with specialization in ìintegrated Circuits and Systems Engineeringî from IIT-Kharagpur, India in Jan Since 2007, he is associated with the Innovation Labs, TCS, Kolkata, working as a senior scientist R&D in the field of embedded signal processing for ubiquitous applications. His overall industry experience is more than 14 years. Prior to that, he has worked with Motorola, India, in the field of embedded signal processing for mobile phones and set-top-boxes. for almost 10 years. He has published more than 15 papers in referred international conferences in which most of them are IEEE sponsored. His current research interest includes context extraction and modeling of user for intelligent human machine interaction. CSI Communications July

14 Research Front Saurav Basu Center for Bioimage Informatics, Carnegie Mellon University Pittsburgh, United States Some Upcoming Challenges in Bioimage Informatics It has been around half a century since the establishment of image processing and computer vision as disciplines firmly rooted in mathematical principles and focused at automatically analyzing visual information in the digital age. The entry into the field of visual automation was both exciting and challenging; problems requiring automation grew exponentially, and so did the gap between arcane mathematical formulae and practical algorithms. It would not be an exaggeration to state that artificial intelligence, machine learning, signal processing, and their direct descendant in the form of computer vision has been instrumental in developing the area of practical and practicable numerical algorithms - mathematics is now a means of livelihood for the computer scientist. Since its inception, there have been numerous directions in which computer vision has taken interesting turns. With the barrage of relevant journals, conferences and workshops, image processing and computer vision has gradually matured into a well-grounded subject that is not merely a collection of ad-hoc solutions for solving academic toy problems, but having the foundation and depth to create a new world of digital assistants. Today, automated algorithms are an integral part in all labor-intensive and observer bias-prone areas such as video surveillance, satellite image analysis, video broadcasting, biological image segmentation, and medical image diagnostics. With the establishment of the indispensability of computer vision algorithms in the age of digital automation and the exponential growth in both the quantity and variability of information, challenges that are unique to this age have emerged which require a long-term evaluation of the future of computer vision....artificial intelligence, machine learning, signal processing, and their direct descendant in the form of computer vision has been instrumental in developing the area of practical and practicable numerical algorithms mathematics is now a means of livelihood for the computer scientist. Application of computer vision algorithms to the gamut of exploratory and diagnostic data in the fields of biology and medicine holds the potential to be one of the greatest success stories in health care and fundamental sciences. Quantitative measurement, computation and informatics from bioimage data, such as brightfield microscopy, confocal microscopy, MRI, CT, and Ultrasound, to name a few, has miles to catch up with the Image acquisition Biologists Noise filtering and reconstruction Signal processors Physicians Image analysis Mathematicians pace of advancement of these modalities. Measurement from biomedical images such as MRI and Ultrasound are fast becoming the primary noninvasive ways for diagnosing abnormalities in patients. Numerical algorithms for segmenting and tracking 3D organ volumes in these images, despite background clutter and modality specific aberration, are the need of the hour for medical diagnostics. The field of microscopy is still a predominantly subjective exercise where qualitative evaluations of the difference between control and experimental groups are the only ways to assess effect of drugs or discover fundamental biological functions - quantitative algorithms that can produce accurate measurements has the potential to advance microscopy in unimaginable ways. Lastly and most importantly, the pace of generation of bioimage data and the availability of physicians and biologists actively participating with the computer vision community means that powerful machine learning algorithms can be Computer scientists Image informatics Statisticians Machine learning/ inference Fig. 1: The conceptual ecosystem of a complete image analysis/computer vision framework devoted to bioimage informatics. The pink rectangles signify the conceptual stages; the gray ellipses show the contribution from different disciplines and communities. The arrows show the direction of travel of knowledge and expertise. CSI Communications July

15 devised to incorporate human experience into a statistical setting in a systematic manner. Fig. 1 shows a version of the current conceptual pipeline of a bioimage informatics system. Here I outline five major areas that I consider to be future challenges that will shape the discipline of automated bioimage informatics in the next age. Open Access to Biological/ Biomedical Data One of the most critical impediments to a computer vision scientist developing specialized algorithms to analyze a particular bioimage segmentation and/or analysis problem is the current disinterest of the biomedical imaging community in establishing an open database of images. Algorithms, often custom-made to solve a particular problem at hand, share a considerable similarity in methodology with other algorithms that might be applied to a different data. As an example, a neuron-tracing algorithm might have the same underlying methodology of matched filtering and background suppression as an actin network measurement algorithm for animal cells. The lack of sufficient open access databases where algorithms can be tested and results reported have led to considerable reinvention of basic methodologies. Different research groups, privy to the data of their own collaborators are often unaware and unsure of the existence of useful algorithms that can already be altered in a minor way to solve their own analysis problems. Moreover, even if a careful study of results reported by different groups in different journals and conferences might lead to an acknowledgement of existing techniques, the lack of open access to the reported datasets is often the reason for the abandonment of those particular methodologies. There is often no way to compare and analyze success of another methodology within the experimental confines of one s own laboratory. Moreover, circumstantial affinity to a distinguished group of biological imagers often puts many computer vision scientists at an advantage compared to equally talented scientists who are not in close collaboration with any facility that can acquire images relevant to the current state of the art. Algorithm developers with lesser resources at hand are often constrained to develop newer algorithms that can only be tested on very generic data that offer no variability or validation for their new algorithms. Quantitative measurement, computation and informatics from bioimage data such as brightfield microscopy, confocal microscopy, MRI, CT, and Ultrasound, to name a few, has miles to catch up with the pace of advancement of these modalities. Therefore, there needs to be a concerted effort by the biological and the biomedical imaging community to develop forums and alliances that can develop open access databases of stateof-the-art images that are in desperate need of quantitative information extraction. Uniform storage, retrieval, and metadata standards need to be decided. There is an immediate need for proactive and useful participation from well-equipped radiology departments of hospitals as well as well funded microscopy laboratories to offer their data to the general computer vision community. It is understandable that privacy and technical concerns will hinder the speedy publishing of data, and in some cases make it perhaps impossible; but for the most part, any disclosure of declassified image data to the general computer vision community will tremendously enhance the streamlined development of the state of the art. Verification and Validation of Algorithms It is undeniable that the numbers of conferences and journals that report current research on bioimage informatics have grown exponentially in the last decade, and so has the reporting of fantastic results. The high frequency of these events has put immense pressure on researchers and reviewers. Several expected incremental improvements over a short period of time naturally have led to the assumption by the reviewers that a huge improvement in the state of the art is a necessity for any consideration for publication. This has naturally resulted in an extremely competitive and claustrophobic atmosphere where there are fewer and fewer incentives for reporting genuine and reasonable results and higher and higher temptations for cherry picking experimental datasets. It is almost next to impossible to reproduce the results reported in a variety of journals without in-depth familiarity with the specific datasets and extensive tuning of free parameters. Methods and numerical analysis often hold for very narrow assumptions and the incentive for the community invariably gravitates towards constructing over complicated and fancy details rather than develop robust, well-rounded, and complete solutions. It is perhaps not uncommon to any of us in the community how a lack of astonishing There is an immediate need for proactive and useful participation from well-equipped radiology departments of hospitals as well as well funded microscopy laboratories to offer their data to the general computer vision community. results often have confined graduate students and lead investigators inside their laboratories for days at end, in search for the magical data or inventing ways to defend comments from reviewers. It is therefore with genuine urgency that the computer vision community, especially those who are engaged in the CSI Communications July

16 field of bioimage informatics, needs to reevaluate and recalibrate the success criteria for the whole community. A more transparent and reproducible model for research needs to be embraced, where the emphasis is less on stellar results and more on the wide applicability, robustness, and reproducibility of research. Higher percentage of numerical and/or qualitative validation metrics needs to be developed, and the focus should shift from vast advancement conditioned on assumptions and constrained data towards robust incremental advancement. A reasonable way to achieve this goal might be for journals to require at least a pseudocode, if not the entire implementation of one s algorithms, with the specification of the exact tuning parameters. Expectations for stellar achievements from the side of the reviewers should be curtailed, and the emphasis should be on usefulness and robustness. This is really the need of the hour - any lack of dedication towards reproducible and incremental but robust results will ultimately lead to a scientific literature that itself generates unhealthy mistrust inside the community. Taking Advantage of the Explosion in Data Acquisition One big change that perhaps sets the current stage apart from the early days of bioimage informatics as well as computer vision is the availability of huge amounts of data generated by experts which stands on the frontier of current human knowledge. High throughput experiments for large biological assays are regular features in the drug research community today. Terabytes of images are generated every day in the pathology research labs for the analysis of cancerous tissues and stockpiling medical evidence of diseases. High-resolution imaging of the neural circuits of Drosophila and the cytoskeletal structure of muscle tissue are carried out routinely throughout universities in the United States that generate gigabytes of information about cellular structure. In short, image data is being generated at a much faster pace than the speed with which computer vision scientists are coming up with tailor-made algorithms that automate specific problems pertaining to each of these datasets. It is therefore quite imperative A more transparent and reproducible model for research needs to be embraced, where the emphasis is less on stellar results and more on the wide applicability, robustness and reproducibility of research. that the process of development of automation algorithms take advantage of this information explosion and modify themselves to be data driven rather than imagination driven. In other words, it would be quite impossible for computer vision scientists to anticipate the physics behind the image generation and then come up with arbitrary and rule-based schemes that work on a wide variation of the images in a particular set. For example, guessing and formulating an efficient stopping force in the case of active contours that segment a large Ultrasound dataset of murine hearts would quite frankly be infeasible - several ad-hoc adjustments need to be carried out to tune and perfect arbitrary and complicated edge stopping formulae. The need of the hour is perhaps to take Group 1 Choose image database Tune algorithm parameters Restricted knowledge sharing Design empirical algorithm Group 2 Choose image database Tune algorithm parameters Restricted knowledge sharing (A) Design empirical algorithm advantage of the immense advancements in the discipline of machine learning and the availability of sufficiently large and annotated datasets by experts in the specific biological problem at hand. In a data-driven atmosphere and with sufficient computing resources, statistical learning models are possibly the best alternatives to learn the intrinsic variability of data. Instead of formulating and anticipating every numerical formula in every algorithm, generic functional forms can be assumed and the corresponding functions learnt as a regression or a classification problem. This learning approach can often lead to very robust and useful solutions in practice that can be easily adapted by practicing biologists and research physicians to actually advance health care and medical diagnostics. Reinforcement learning and online learning models can be easily incorporated in a rapidly functioning image generation laboratory that can quickly adapt to new annotated data. With the current capabilities in computing speed and power for most computer vision groups, and availability of collaboration between the medical and the vision communities, incorporating human knowledge and driving the solution process through the data is the most natural choice in today s world. Dissemination of Diagnostic Data for the End User A very reasonable and valid impetus for the continued relevance and usefulness of the computer vision community in the field of bioimage informatics is the Open access image database Data-driven algorithm generation Standardized and reasonable validation methodologies Provide fast diagnostic software to end-user Group 1 Group 1 (B) Group 3 Fig. 2: A schematic comparison between current and desired practice of bioimage informatics. (A) shows the isolated generation of numerical algorithms by several independent groups that leads to reinvention and lack of practicability (B) shows an idealized cooperation between groups with free information exchange and use of published databases and validation technologies. CSI Communications July

17 ultimate utilization of fast diagnostic algorithms by the end user. Although part of the importance of the community is discovering important facts in the fields of clinical research and fundamental science, yet, the biggest service and impact will indeed come from making vision algorithms operate under limited resources inside the house of the common man. Quick and accurate medical diagnosis must be performed on A focused curriculum on bioimage analysis and informatics needs to be developed that can enable the undergraduate student to bridge the daunting gap between engineering and biology. handheld medical devices such as a heart Ultrasound device and remedial measures advised on the fly. This goal can only be achieved if simple, fast, and accurate algorithms are developed that can run on limited resources and computing power. The huge array of statistical models and trained examples can be suitably accessed in a client server model where the client only performs a regression, classification, or computation task and the bulk of the expertise is transferred through the server which has the benefits of high computational power and accumulation of expertise in specialized laboratories. In short, in addition to developing dataintensive and mathematically complete algorithms that can solve complicated problems in a high-tech environment, low-tech versions of the same algorithms need to be developed that disseminate the same fruits of success to the ultimate end user and can be run via accessing the information pool through the Internet. This would ensure that the science of computer vision makes the greatest impact on everyday life. Fig. 2 contrasts the way in which the aforementioned four points are implemented in the current scenario versus an idealized way the algorithm generation community might work in order to reduce reinvention and increase practicability of academic inventions. Education and Training of the Modern Generation of Biomedical Image Analysts I come to the last and most frequently overlooked aspect of the recipe to success of bioimage informatics as a discipline. Education and hands-on training of the modern generation of computer vision scientists with a specific focus on bioimage informatics is in a state of confusion and disarray. Biomedical engineering is in most universities a newly formed discipline at the undergraduate level. A subject-specific curriculum and philosophy has still not permeated the classrooms from the highly specialized laboratories that perform cutting edge research on bioimage informatics. Successful research in vision algorithms applied to biomedical images have frequently taken years of study for most investigators in various disciplines ranging from mathematics, signal processing, and biology - this firmly establishes the credentials of bioimage informatics as a multidisciplinary subject that not only requires perspective from different angles and disciplines, but a need to develop an attitude for collaboration and special understanding for cross-discipline communication. Often, the demands from the biology community seem imprecise and overcomplicated to the computer vision community, and the mathematical jargon and computations supplied by the computer vision community seem unnecessary and impractical to the biology community. The need to develop a language that can enable effective and free communication is therefore paramount. And it is only natural that experts in bioimage informatics need to take the initiative to develop this language and disseminate this perspective through undergraduate education. Undergraduate curriculums in bioimage analysis and informatics not only need to teach the systems perspective of signal processing and abstraction of real medical imaging devices as systems, but also need to entrench the student in strong mathematical principles of algebra and analysis. Moreover, a systems-oriented biology syllabi and current understanding of biological and medical imaging research that can lead to fundamental discoveries is also a unique requirement to this field. Simply put, one cannot depend on years of practical research experience to develop a bioimage informatics expert - a focused curriculum on bioimage analysis and informatics needs to be developed that can enable the undergraduate student to bridge the daunting gap between engineering and biology. n About the Author Saurav Basu is currently engaged in developing bioimage informatics retrieval systems in the Center for Bioimage Informatics, Carnegie Mellon University, USA. He was conferred a PhD in Image Analysis by the University of Virginia, USA, in 2011, where he worked with Prof. Scott T. Acton to develop mathematical algorithms for image recognition that leveraged techniques in differential geometry. He has been actively involved in the image analysis community as a researcher and reviewer for distinguished journals and conferences. His research interests include application of differential geometry, graph methods, machine learning, and PDEs to problems in biomedical image registration, distance metric calculation, and generative modeling of biological data. CSI Communications July

18 Research Front R P Ramkumar* and Dr. S Arumugam** *Research Scholar, Mahendra Institute of Technology, Namakkal id: rprkvishnu@gmail.com **Chief Executive Officer, Nandha Educational Institutions, Erode id: arumugamdote@yahoo.co.in Accurate Pupil and Iris Localization using Reverse Function Abstract: The goal of iris recognition is to recognize human identity through the textural characteristics of one s iris muscular patterns. The iris recognition has been acknowledged as one of the most accurate biometric modalities because of its high-recognition rate. In order to provide competent and successful iris recognition, iris localization plays a major role. If the inner boundary and outer boundary of iris are detected accurately, then the efficiency of the subsequent stages will be increased and also the computational cost will be reduced. In this proposed iris localization method, pupil localization is done by using scaling, reverse function, and four neighbors method so that irrespective of pupil s contour, either circle or ellipse, the pupil s boundary is detected accurately. For iris outer boundary detection, contrast enhancement, special wedges, and thresholding techniques are used to isolate the specific iris regions without eyelid and eyelash occlusions. Upon completing the above phases, pupil boundary is detected 100% perfectly. Keywords: iris recognition, iris localization, iris segmentation, thresholding, neighborhood method, contrast enhancement Introduction The process of iris recognition consists of subcomponents, viz. iris localization, iris segmentation, normalization, iris feature extraction, and matching. Since the recognition accuracy and efficiency mainly depends on iris feature extraction and classification of discriminant features, and in turn these stages are dependent on iris localization. This paper mainly focuses on iris localization process. Section 2 deals with overview of existing iris localization algorithms. Then Section 3 insists on the proposed method. Section 4 imparts with experimental results and Section 5 draws the conclusion. Comparison of Existing Algorithms Fast Localization Method Iris localization deals with the detection of outer boundary (iris/sclera) and inner boundary (iris/pupil). Classical approach deals with two steps, viz., edge point detection and Circular Hough Transform (CHT). In this approach [17], in order to reduce the searching complexity, Bounding Box (BB) is introduced. Steps for outer boundary detection include: (a) Horizontal Band (HB), which is defined from middle row to one-fourth height of image; (b) HB is binarized by using proper threshold value; (c) BB is defined for pixel intensities within the range of (m-2σ, m+2σ), where m is mean and σ is standard deviation; and (d) First Canny edge detection operation is applied for BB region and subsequently CHT is applied by ignoring smaller edge components, which results in outer boundary detection. Steps for inner boundary detection include; (a) Average intensity of the iris region pixels are calculated and it is considered as new threshold value th ; (b) If the pixel (x, y) has intensity less than th value and also its distance from the centre of outer circle (x 0, y 0 ) is less than half of the radius r of the outer circle, then this pixel belongs to pupil area, and this will lead to define the BB; and (c) Canny edge detection process is applied for this BB region and subsequently CHT is also applied, and by filtering some of the smaller edge components also lead to inner boundary detection. Experimental results with the UBIRIS database [20] show that, execution time and accuracy is very much improved for iris localization process when compared with Wilde s approach [37], and hence results in fast method for iris localization. Sclera Iris Inner boundary Pupil center (c 1,c 2 ) l 1 l 2 l 4 Iris center (c 3,c 4 ) l 3 θ 2 θ 1 Pupil Outer boundary Fig. 1: Ten parameters of deformable iris model Phase-based Iris Recognition Algorithm Efficient fingerprint algorithm using phase-based image matching is already implemented by these authors in [18], and now they have mobilized the same principle (use of Fourier phase information of iris image) for Iris recognition too [18]. With respect to Iris localization process, images are first converted into gray scale image. Fig. 1 shows ten parameters of deformable iris model used in this system. Steps involved in inner (iris/pupil) boundary detection are: (a) The circumference of inner boundary is considered as ellipse, with two principle axis (l 1, l 2 ), center (c 1, c 2 ), and the rotation angle θ and (b) The optimal estimate (l 1, l 2, c 1, c 2, θ 1 ) is found by maximizing the value of s(l 1 + l 1, l 2 + l 2, c 1, c 2, θ 1 ) - s(l 1, l 2, c 1, c 2, θ 1 ), where l 1 and l 2 are small constants, and s denotes the N-point contour summation of pixel values along the ellipse and is defined as: where, p 1 (n) = l 1 cosθ 1. cos((2π/n)n) l 2 sinθ 1. sin((2π/n)n) + c 1, p 2 (n) = l 1 sinθ 1. cos((2π/n)n) + l 2 cosθ 1. sin((2π/n)n) + c 2, f org is the original image. So that, the inner boundary is detected, where there will be a sudden change in luminance summed around its perimeter. Similarly, the optimal estimate (l 3, l 4, c 3, c 4, θ 2 ) for the outer boundary is detected with the path of contour summation changed from ellipse to circle (i.e., l 3 = l 4 ). In the normalization process, unwrapping of iris region to rectangular block of fixed size 256 x 128 pixels is done for only lower half of iris image. By using the local Histogram Equalization (HE) technique, the contrast enhancement is made. For matching process, principle of phase-based image matching by using the Phase-Only Correlation (POC) function is employed. Improved Iris Segmentation Algorithm In this approach [14], reimplementation of Daugman s algorithm [9] developed by Masek [16], to yield the improved segmentation process is carried out and this is called as ND_IRIS. Here, iris recognition process consists of three phases, viz. iris segmentation, iris encoding, and iris matching. Second and third phases are same as Masek s approach. In segmentation phase for generating the edge map, Canny edge detector and CHT are employed. Hough space s maximum value gives the center and radius of circle. CSI Communications July

19 Steps in optimization process includes: (a) Reverse the detection order: With respect to Masek s algorithm, at the beginning, outer boundary of iris is detected and then the inner iris boundary is detected. Here this process is reversed, i.e. pupil boundary (inner) is detected first and the iris boundary (outer) is detected using Hough transform. (b) Reduce edge points: In order to improve the generated results, pixels with more than high-intensity value (240) and pixels below the low-intensity value (30) is removed, so that boundary is identified clearly. (c) Modification to Hough Transform: In Masek s approach, the Hough Transform votes for center location in all directions for each edge point and a given radius r. But in this approach, each edge point votes for possible center locations in the area with in only 30 on each side of local normal directions. (d) Hypothesize and verify: In Masek s approach, edge pixels are detected by using Canny Edge Detection method and then peaks in Hough Space is used to detect the boundaries in the eye image. In this approach, a test is performed with the peaks in Hough space and this check is done on both (left and right) sides of two boundaries (limbic and pupillary) respectively. Moreover, additional check is also done to ensure that iris-pupil boundary should be within the region of sclerairis boundary, and the centers of two circular boundaries should be closer than half of the radius of iris-pupil boundary. (e) Segmentation improvements: Due to implementation of previous stages, ND_IRIS results in improved segmentation process when compared with Masek s approach. During some specific circumstances, where the image with very low contrast and more occlusions, ND_IRIS also leads to improper segmentation. (f) Eyelid detection: In the Masek s algorithm, eyelids are considered as horizontal, lines using linear edge maps are generated using the Canny edge detector. Whereas in this approach, each eyelid is considered as two straight lines. After this iris image is divided into four equal parts of equal size. Now eyelid is detected in each of the four windows and resulting images are combined together to yield the refined and segmented iris from eyelid occlusions. In the normalization process, the unwrapping of iris image takes place and for encoding Gabor filters are used. For matching process, Hamming Distance (HD) is employed. Overall experimental result shows that, this approach leads to an increase of about 6% than Masek s segmentation method by using the rank-one recognition rate. Using Maximum Rectangular Region for Iris Recognition In this approach [22], large area axis-parallel rectangular Region of Interest (RoI) is extracted from normalized iris image so that success rate of this algorithm is very much increased. Steps for iris localization includes: (a) Iris and pupil boundaries are considered as circles; however they are not concentric; (b) At the beginning, rough estimation of pupil center (x p, y p ) is determined by projecting the image in vertical and horizontal directions; (c) By choosing a proper threshold value, and keeping this coordinate (x p, y p ) as center, binarization is done; and (d) More accurate pupil coordinates are obtained with the resulting binary region. Afterwards Canny edge detection and Hough transform is applied to get the exact pupil center (x p0, y p0 ), pupil radius r p, iris center (x i0, y i0 ), and iris radius r i. In the normalization process, unwrapping of iris image takes place; subsequently Histogram Equalization (HE) is used for image enhancement. Two stages, viz. RoI and feature coding (with the help of Cumulative-sum-based analysis) are used for feature extraction and HD is used for matching process. By choosing a proper threshold value(s), this algorithm furnishes 98.37% efficiency, which is far better than the algorithms described in [4,15]. Feature Extraction after Texture analysis In this approach [3], after acquiring the color iris image, it is processed to grayscale image. By using a specialized algorithm called Feature Extraction Algorithm, the IRIS Effective Region (IER) is detected and extracted from the grayscale iris image. Based on the degree of similarity between the presented iris image and stored iris image, corresponding individual is identified or rejected. Conversion from 84-bit BMP color image to 8-bit grayscale image includes: (a) Input image: 24-bit BMP iris image of six 100*100; (b) Convert the RGB value to Gray values using the following notation, bv = * rv gv * bv, gv = * rv gv * bv, rv = * rv gv * bv, grv = bv = gv = rv, where bv = blue value, gv = green value, rv = red value, grv = gray value; and (c) The result is written into new BMP file which is 8-bit gray scale image. Accomplishing IRIS Edge Detection includes: (a) Convert the 8-bit gray scale iris image into planar image; (b) A 3 x 3 float type Horizontal and Vertical kernels are set as: (c) Planar image (created in step a) is passed through the kernels (created in step b); and (d) Modified fine-grained planar image is stored. Output Image at this stage will have distinct marks in distinct iris areas, thereby the edge detection process is carried [3]. After this, IER of size 8 x 12 and n number of similar IERs are extracted, which is used for effective individual identification. Based on Using Ridgelet and Curvelet Transform In this approach [19], after segmentation and normalization process, a new method is employed for feature extraction with the help of Ridgelet and Curvelet transform. Steps for Segmentation includes: (a) Inner and outer boundaries of iris regions are first identified approximately by using Canny method, then thresholding is made with the 3 3 window; (b) Now to detect the accurate boundaries, CHT is used with the predefined radius for segmenting the iris region; and (c) Here the exact boundary of iris is not determined, instead with the predefined radius from the center of pupil itself, the regions very close to pupil called collarette portions alone are considered. This leads to 98.64% of accuracy in iris boundary detection and it is high when compared with algorithms described in. Steps for Normalization and Enhancement include: (a) Daugman s Rubber Sheet Model is employed and the rectangular window of size is chosen such that it is free from eyelids and eyelash occlusions and (b) In order to identify the distinguished portions of iris region, contrast enhancement of image becomes necessary. Therefore for enhancing the image, median filter, HE, and 2D Wiener filter is used. Here Ridgelet and Curvelet transforms are employed to extract the distinguished features and HD is used for matching process. When compared with various existing algorithms, this approach with Ridgelet transformation gives 97.96% and Curvelet transformation gives 98% accuracy. Based on DWT and Haar Wavelets A robust iris recognition algorithm using 2D DWT with Haar Wavelet is implemented CSI Communications July

20 in this approach [7]. Here for detecting and segmenting the boundaries of iris and pupil, Integro Differential Operator (IDO) is used, followed by Daugman s Rubber Sheet Model for normalization process. For feature extraction process, 2D DWT with Haar Wavelet is applied up to 4th and 5th level of decomposition. From this level, equivalent binary codes are generated. For matching purpose, HD is used to find the dissimilarity ratio between iris codes. The performance measure of this approach is given by FRR with 0.33%, when compared with VeriEye Algorithm [21] whose FFR is 0.32% and FAR is 0.001%. Efficient Iris Localization Algorithm In this approach [10], mainly three-level thresholding and morphological processing is carried out for quick processing of iris localization process with acceptable accuracy. This approach has two stages, viz. Coarse and Fine stage. Steps in coarse stage includes: (a) Input image is processed with Gray-scale closing and then three-level thresholding is chosen for bright, medium, and dark intensity pixels respectively to yield binary image and (b) Morphological processing is done to fill the small holes so that the pupil alone is detected correctly and approximate center of pupil is also marked. Steps in fine refinement stage includes: (a) Image size is reduced to one-fourth of its size, so that the search region is also reduced; (b) With the help of neighborhood pixels and the approximate center from the coarse stage, apply the Daugman s IDO to detect the pupil and iris boundaries and subsequently its centers are also identified; (c) With respect to iris boundary detection process, instead of searching the entire boundary, the sectors ranging from +30 to -30 on both sides (left and right) from the initial center alone is considered so that occlusions due to eyelids and eyelashes are very much reduced; and (d) Now these two circles with their boundaries and centers are superimposed on the original iris image, so that iris region is isolated perfectly. CASIA V3 database [6] is used for evaluating purpose with 1817 iris images. When this approach is compared with Masek s approach [16], time taken for iris localization is 20 sec for Masek s with accuracy of 87%, whereas this approach yields 98.8% accuracy with duration of 240msec only. Proposed Method In order to have an efficient iris recognition algorithm, the noises in the acquired iris image such as eyelids, eyelashes, and reflections due to lighting effect should be removed completely or reduced to some extent. Therefore the iris localization process plays a crucial role, since the left behind processes (such as normalization, feature extraction, and matching) directly depends on the segmented iris regions. The proposed method include two steps, viz. pupil localization and iris localization respectively. Steps for Pupil localization In order to detect the pupil boundary three phases are used, viz. scaling, reverse function, and four neighbors method respectively. The scaling phase (first phase) deals with reducing the iris image size into one-fourth of its original size, so that the region for thresholding process is very much diminished, which in turn reduces the processing time and computational cost, so that it paves path for increasing the accuracy and efficiency of the entire iris recognition process. Since the pupil is the darkest area in the entire eye image, this portion is distinguished by thresholding operation with the help of reverse function as shown in Fig. 2(a) and 2(b). This constitutes the second phase. (a) (b) (c) Fig. 2: (a). Original image, (b) After applying negative function, and (c) Image after segmentation As a result of this, pupil region is detected approximately and this constitutes the first phase of pupil localization stage. In the third phase, four neighbors method based on [1-2] is applied for each pixel as shown in Table 1, to localize the pupil boundary accurately. X (x - 1, y) X (x, y - 1) (x, y) (x, y + 1) X (x + 1, y) X Table 1: Four neighbor method where, (x-1, y) represents top most point, (x+1, y) represents bottom most point, (x, y-1) represents left most point and (x, y+1) represents right point most in pupil boundaries respectively. Now, the pixel values are checked such that, if its value is equal to white and one of its four neighbors value is less than the white color, then corresponding pixel value is replaced by white color else original eye images pixel value is retained. Upon completing the above three phases, pupil boundary is detected accurately as shown in Fig. 2(c), irrespective of its shape either circle or ellipse, with an average processing time of only 0.7 to 0.8 seconds. Steps for Iris localization Here three steps, viz. contrast enhancement, special wedges, and thresholding are employed to detect the specific iris regions from an eye image. The dedicated algorithm described in [11] is used for contrast enhancement, so that limbic boundary is identified clearly as shown in Fig. 3(b), when compared with the original image shown in Fig. 3(a). (a) (b) Fig. 3: Eye image before and after enhancement process Now, the pixels lying only within the region of ±45 along the central axis on both sides, i.e. left and right sides of iris regions, alone are considered as illustrated in [12] is shown in Fig. 4, which is free from eyelash and eyelid occlusions Central axis Fig. 4: Measured regions of arcs in the iris region along the central axis This portion of measured iris region is the maximum useful region with minimum noise and is said to be the Region of Interest (RoI), and its mean value is also calculated. By choosing the suitable and predetermined threshold value below the mean value, arc regions of iris alone are isolated and hence the process of iris segmentation from limbic boundary is accomplished perfectly. Experimental Results This proposed method is tested on CASIA iris image database from Institute of Automation, Chinese Academy of Science (CASIA) [5], with 108 different subjects. With this algorithm, iris localization and iris segmentation processes are done exactly and accurately, even though some of the iris images are occluded by eyelids and eyelashes. The performance measures are compared with some of the existing algorithms for iris localization is shown in Table 2 is depicted from [25]. From the results CSI Communications July

21 shown in Table 2, this proposed method has enhanced performance. By using this proposed method, the pupil and iris segmentation is done accurately. Moreover this method compensates for head tilt, pupil size, pupil shape (i.e. either ellipse or circle) and also in terms of reduced computational cost. If the acquired image itself is not clear due to noises and it does not have distinct sclera boundary even after images contrast enhancement process, then it will not be recognized correctly. During these situations, this proposed algorithm has degraded performance. Methods Recognition Rate (%) Daugman [9] 98.60% Wildes [23] 99.50% Cui [8] 99.30% Maryam [24] 99.28% Jarjes [13] 98.85% Sundaram [17] 98.43% Liu [14] 97.08% Essam [10] 98.80% Proposed Almost 100% Table 2: Comparison of Algorithms Conclusion Out of various iris recognition stages, viz. iris localization, iris segmentation, normalization, iris feature extraction and matching process, pupil and iris localization process plays a vital role, because its accuracy determines the overall iris recognition algorithm s efficiency. When compared with various exiting algorithms, this proposed method grants faultless identification and accurate separation of pupil and iris boundaries, irrespective of its shape either ellipse or circle, to almost 100% absolutely with less computational cost. Acknowledgment In this paper, as a part of the research, the iris database of version 1 is used and is collected by the Institute of Automation, Chinese Academy of Sciences (CASIA). References [1] Almisreb, A A, et al. (2010). Pupil Localization Using Negative Function and the Four Neighbors. Second International Conference on Computational Intelligence, Modelling and Simulation, [2] Almisreb, A A, et al. (2011). Enhancement Pupil Isolation Method in Iris Recognition. IEEE International Conference on System Engineering and Technology (ICSET), 1-4. [3] Bhattacharyya, D, et al. (2008). IRIS Texture Analysis and Feature Extraction for Biometric Pattern Recognition. International Journal of Database Theory and Application, 1(1), [4] Boles, W W and Boashsh, B (1998). A Human Identification Technique Using Images of the iris and Wavelet Transform. IEEE Transactions on Signal Processing, 46(4), [5] CASIA, Iris Image Database, sinobiometrics.com [6] CASIA-IrisV3 Database [Online]. Available: asp [7] Chirchi, V R E, et al. (2011). Iris Biometric Recognition for Person Identification in Security Systems. International Journal of Computer Applications ( ), 24(9), 1-6. [8] Cui, J, et al. (2004). A fast and robust iris localization method based on texture segmentation. Proceedings of SPIE, 5404, [9] Daugman, J (2004). How iris recognition works. IEEE Transactions on Circuits and Systems for Video Technology, 14(1), [10] Essam, M, et al. (2012). C16. An Efficient Iris Localization Algorithm. 29th National Radio Science Conference (NSRC 2012), Cairo University, Egypt, [11] Hong, L, et al. (1998). Fingerprint image enhancement algorithm and performance evaluation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 20(8), [12] Horapong, K, et al. (2005). An Iris Verification Using Edge Detection. ICICS2005, [13] Jarjes, A A, et al. (2010). Iris Localization: Detecting Accurate Pupil Contour and Localizing Limbus Boundary. Second International Asia Conference on Informatics in Control, Automation and Robotics, [14] Liu, X, et al. (2005). Experiments with an improved iris segmentation algorithm. Fourth IEEE Workshop on Automatic Identification Advanced Technologies, [15] Ma, L, et al. (2003). Personal Identification Based on Iris Texture Analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(12), [16] Masek, L (2003). Recognition of Human Iris Patterns for Biometric Identification. The University of Western Australia, csse.uwa.edu.au/~pk/studentprojects/libor/ [17] Meenakshi Sundaram, R, et al. (2011). A Fast Method for Iris Localization. Second International Conference on Emerging Applications of Information Technology (EAIT), [18] Miyazawa, K, et al. (2005). A Phase-Based Iris Recognition Algorithm. LNCS 3832, [19] Najafi, M and Ghofrani, S (2011). Iris Recognition Based on Using Ridgelet and Curvelet Transform. International Journal of Signal Processing, Image Processing and Pattern Recognition, 4(2), [20] Proena, H and Alexandre, L (2005). UBIRIS: A noisy iris image database. 13th International Conference on Image Analysis and Processing (ICIAP2005), Vol. LNCS 3617, Springer, [21] Verieye iris recognition concept for performance evaluation, [22] Viriri, S and Tapamo, J-R (2009). Improving Irisbased Personal Identification using Maximum Rectangular Region Detection. International Conference on Digital Image Processing, [23] Wildes, R P (1997). Iris recognition: an emerging biometric technology. Proceedings of IEEE, 85, [24] Yazdanpanah, M and Amini, E (2009). Fast Iris Localization in Recognition Systems. International Instrumentation and Measurement Technology Conference (I2MTC09), [25] Ziauddin, S and Dailey, M N (2009). A Robust Hybrid Iris Localization Technique. 6th International Conference on Electrical Engineering/ Electronics, Computer, Telecommunications and Information Technology, (ECTI-CON 2009), 2, n About the Authors R P Ramkumar is working as Assistant Professor at Mahendra Institute of Technology, Namakkal. He has done B.E. and M.E. in Computer Science & Engineering. Currently, he is pursuing PhD in the field of Biometrics. His area of interest includes image processing, pattern recognition, and information retrieval. He can be reached at rprkvishnu@ gmail.com. Dr. S Arumugam is presently working as Chief Executive Officer, Nandha Educational Institutions, Erode. He has done B.E. in Electrical Engineering and MSc (Engg.) in Applied Electronics from P. S. G. College of Technology, Coimbatore, University of Madras. He obtained his PhD degree in Computer Science & Engineering from Anna University in He worked in the Directorate of Technical Education and retired as Additional Director of Technical Education in He has published more than 100 papers and guided 15 PhD scholars. He has membership in IEEE, FIETE, FIE (I), SMCSI, & LISTE. CSI Communications July

22 Article Hema Ramachandran Speed-IT Research Fellow, College of Engineering, Trivandrum Image and Video Processing Toolbox in Scilab SCILAB is a free software alternative to MATLAB, hailed very often as the language of technical and scientific computing. MATLAB has found a permanent place not only in the curriculum of applied science and engineering studies, but also in research and development arena. SCILAB closely mimics functionalities of MATLAB and is emerging as an effective alternative. Scilab has a rich collection of tool boxes suited for applications in science and technology fields. The SIVP toolbox is an image and video processing toolbox, which supports formats like BMP, PNG, JPEG, TIFF, PBM, PGM, PPM, and SR. It can do a variety of applications like video I/O, camera read, image type conversion, spatial transformation functions, image analysis and statistical functions, image arithmetic functions, linear filtering, morphological operations, and color space conversions. The Scilab Image and Video Processing (SVIP) toolbox can be downloaded from the website address sourceforge.net. Though SIVP handles the following types of image files: JPG, BMP, PNG, TIFF, and a few others, as JPG being the most popular one, we will confine our examples to this file type. It may also be noted here that JPG compresses the image data and stores it in a compressed format. The data is uncompressed before using this file. These are of course transparent to the Scilab user. For all our examples, we will use a color jpg file kavya.jpg. We assume that our test file is available on the default directory (verify your default directory using pwd). We will first see how an image file is opened and viewed. This is rather straight forward. x=imread( kavya.jpg ); imshow(x); Fig. 1: Showing an image We can now view the image data and try to correlate with this image. We can directly inspect the array x, as follows: Let us work with a gray image. Scilab lets you convert RGB image to gray image with rgb2gray() command. x=imread( kavya.jpg ); y=rgb2gray(x); imshow(y); Fig. 3: Gray image After making any change in an image, it can be written into a file under any name using the imwrite() command. We will save the gray image of kavya in kavya1.jpg using the command imwrite(y, kavya1. jpg ). We can also try a reverse activity - that of converting an ordinary matrix into an image. A matrix can be converted into an image with the mat2gray() command. for i=1:100 for j=1:100 x(i,j)=i+j; end; end; y=mat2gray(x); imshow(y) Fig. 2: Displaying image data As an image is nothing but an array of numbers, we can do any arithmetic operations on images. Some of these Fig. 4: Converting a matrix into an image operations give useful effects and some are of mere curiosity. Let us see a particular case of subtraction. Resize your working image to size 100*100 before you start. Kavya2.jpg has already been resized to this size. Now let us find out the maximum intensity in the file. x=imread( kavya2.jpg ); max(x); 243 Now we can create an array y of the same size as the image with all elements equal to the maximum intensity, and then subtract x from y: y=243*ones(100,100) z=y-x; imshow(z); CSI Communications July

23 Fig. 5: Negative of an image We can see the negative of the original image. The arithmetic behind the process is subtraction of all intensity values from the maximum. There are commands for addition, subtraction, multiplication, and division of images: imadd(), imsubtract(), immultiply(), imdivide(). These can take images as arguments. In this case, they should both be of the same size. The second argument can also be a scalar. Subtraction is of special use in comparing two images with small differences. The subtraction will indicate the difference alone. Open kavya.jpg in an editor such as Paint and make a small change to add a detail (you may do in the color image and then convert into gray using Scilab). Here we have added a beauty spot. Fig. 6: Image opened in Paint and edited If kavya2.jpg and kavya with spot.jpg are the two images (both having been converted to gray), then see the effect of the subtraction below: x=imread( kavya2.jpg ); y=imread( kavyawithspot.jpg ); z=imsubtract(x,y); imshow(z); Fig. 7: Image subtraction We will now introduce imhist( ) command that computes the histogram of intensity distributions in an image. This can be used for studying contrast of images and then adjusting it favorably. Let us see an example of computing and plotting histogram. x=imread( kavya2.jpg ); [p, i]=imhist(x) //p=pixel count and i= intensity plot(i,p) Fig. 8: Histogram of an image The intensity histogram of an image is the plot of count of pixels in each intensity level. For example, in the histogram above, the maximum count is 100 at the intensity 150 (approximately). Image processing has three classes of operations: point, local, and global. Point operations process each pixel independent of any other. Example is negativing. Global operations consider some global features of an image and modify every pixel based on it. The histogram is a global measure. We will now acquaint ourselves with the remaining class - local. These are by and large known as filtering and consider the neighboring pixels of each pixel to modify it. Let us first see a simple example before we look at shorthand Scilab commands. A popular filtering known as Blurring Filter simple averages nearby pixel intensities. This makes the picture less sharp. For example, pixel x(5,5) is replaced with average of x(5,4), x(5,5), and x(5, 6) - the left, current, and right. Leaving out left and right border (which have respectively no left and right!), we can do this with a simple program: x=imread( kavya2.jpg ); for i=2:99 for j=2:99 y(i,j)=(x(i,j-1)+x(i,j)+x(i,j+1))/3 end end y=mat2rgb(y) imshow(y) The above operation can be compactly written using a matrix. Such matrices are called filter kernel matrices. In the above case, the filter kernel is: F= [ /3 1/3 1/ ] If we place this matrix over any area of an image matrix, 9 intensity values will be there corresponding to each kernel matrix element. If we multiply these corresponding values and add, we can see the same effect that our original example had achieved. Consider the following image matrix: If we place the kernel matrix at the top left corner, we get the following correspondence: / / / CSI Communications July

24 If we multiply each pair and add, we get 160/3+200/3+201/3 = 173. We now place 173 in place of the central position of the kernel, the position of 200, in a new image matrix. Now we shift the kernel one position to the right and repeat the process and we get a replacement for 210. After we finish the first row, we will start again from the left, but this time, one position down. We will now start replacing 173, It is obvious that the border elements will not get altered by this process. Most filters leave the border elements untouched and in a big image, it would not be noticeable. Most filtering processes can be boiled down to a filter kernel matrix such as the above. Once the matrix is available, filtering can be achieved with the imfilter ( ) command. The above example can be implemented as follows: x=imread( kavya2.jpg ); F= [ ]; y=imfilter(x,f); imshow(y) Fig. 9: A low-pass filter The effect of this filtering is not very prominent. A stronger averaging will produce striking effect. Let us consider a filter that represents averaging of 25 pixels in a grid of 5 5 size. This can be represented by a matrix of 5 5 size with each value = 1/25 = 0.04 as follows: F= [ ]; Of course, this can be created easily in Scilab in a quicker way and be used to filter an image: F= 0.04*ones(5,5); y=imfilter(x,f); imshow(y) Fig. 10: A strong low-pass filter Now the effect of blurring is very prominent. This filter is also known as low-pass filter as it lets low frequency (small intensity variations) to pass and evens out high frequency (sharp intensity variations). The exact reverse of this is the case with high-pass filters, which are ideal for detecting edges in images. A simple high-pass filter is given by the Kernel: F= [ ] (The curious reader can try applying this on a small matrix of numbers to verify that it has no effect if all the numbers in the matrix are the same, whatever be their value. If the matrix had all zeroes in first three columns and all 100 in next three columns, then the filter highlights only the crossover area). Let us now use the filter: x=imread( kavya2.jpg ); F= [ ] y=imfilter(x,f); imshow(y) Fig. 11: A high-pass filter A stronger filtering is possible with a 9 x 9 filter with all values -1, except the central value, which is 81. x=imread( kavya2.jpg ); F=-1*ones(9,9); F(5,5)=81; y=imfilter(x,f); imshow(y) Fig. 12: A strong high-pass filter Thus, we see that a number of visual effects can be produced using filtering and the same can be condensed to simple arithmetic operations. A bidirectional association between visual effects and arithmetic operations is thus enabled through these experiments. n About the Author Hema Ramachandran is Speed-IT Research fellow at the College of Engineering, Trivandrum. She was formerly principal of the University College of Engineering, Karyavattom. She holds B.Tech, M.Tech degrees in Electrical Engineering and M.Phil degree in Futures Studies and also a PG diploma in Software Engineering. She has taught in various engineering colleges and technical institutions in India and abroad. She had a brief stint in the IT industry as well. Her current area of interest is Wireless Electricity. She is author of a book on Scilab published by S. Chand Publishers, New Delhi. CSI Communications July

25 Article Dr. Pramod Koparkar Senior Consultant Importance of Shifting Focus in Solving Problems This article describes a case study illustrating how techniques from one discipline (Image Synthesis) can be applied to solve problem in other discipline (Geometric Modelling). Once the focus from the main problem is shifted to the user requirement analysis, it leads its own natural path to the solution. The Problem at Hand This is all about designing cars and their front mudguards (and modelling them in a computer using software). A tire of a car is a simple concept that has helped to make our lives more enjoyable, more comfortable, and safer. The design of a tire has evolved along with the design of a car over last century or so. The increasing speed of cars has put up many constraints on both. The high speed of a car requires that (1) the tires should be bigger and flatter/ broader, and (2) the engine should be more powerful, and hence, bigger in size. Both of these lead to a requirement of reducing wastage in space while designing the car. The mudguard should be as small as possible to accommodate a larger engine, but at the same time, it should be as big as possible to accommodate larger tires. A degree of difficulty is added by two more observations: (1) the front wheels (and their tires) turn in different orientations from left to right, and (2) different tire manufacturers offer a variety in tire types, and the end user can choose any of them. All possible orientations of all possible types of tires together would occupy (or designate) some solid volume in space. One needs to model its outermost encompassing surface, known as the Minimal Enclosing Envelope. Just visualize it as some kind of elastic bag or balloon containing the designated volume and enclosing it tightly. Once the envelope is established, the mudguard can be easily modeled using offset surface of this envelope. Thus, any solution to design the mudguard essentially boils down to the solution of designing the envelope. The Complexity of the Problem In 3-D Geometric Modelling, often some kind of approximation is used. A curved surface may be represented by a collection of very small triangles fitting edge-to-edge and not deviating too much from the original surface [1]. These serve a good (within tolerance) representation. A single tire takes around 30,000 such triangles to be within manufacturer s tolerance. We are given these triangles (x-y-z coordinates) in their basic or canonical form. The rear tires remain in their canonical form, while the front tires turn and change in orientation. The orientation angles can be approximated using around 128 different angular values. One needs to calculate the triangles (x-y-z coordinates) in each of this orientation. Additionally, there are around 20 different types of the tires per single model of a car. Thus, we are lead to deal with 30,000 X 128 X 20 = 60,000 X 128 = 76,800,000 (approximately 80 Million) triangles while deciding on their envelope. Finding out the Minimal Envelope requires sorting the triangular data. This is typically an O(n 2 ) algorithm where n is the number of items to be sorted [2]. In our case n is 80,000,000 and n 2 is 6,400,000,000,000,000 or 6,400 Trillion! Expectations from the Solution Geometric Modeler is a software program to create models and to operate on them in order to answer various questions about them [3]. Even a good geometric modeler on an efficient (speed-wise) and large (memory-wise) machine requires a lot of time to solve this mudguard problem. When this problem caught my attention, the calculation was typically taking hours (yes, hours!) to generate and view the mudguard. This was not a very great show: Feedback after hours was definitely hampering the designer s thinking/ creative process. A designer would suggest some design and input its required parameters. Then goes home to come back next morning to see what has happened, how the mudguard now looks. If not satisfactory, again tries changing some parameters, and goes home while the machine is taking next hours to do the job! Again, if there is some modification Well, you can now see! The design life cycle typically requires fast see-modify-see-modify-see and the response time must ideally be in seconds. The designers may compromise for a few (typically 2-3) minutes, but hours are too much! The Crux of the Problem Creating an envelope enclosing 80 million triangles is a 3-D problem. Its complete solution as per Geometric Modelling standard does require sorting each triangle against every other triangle. Of course, some saving can always be done using data coherence and spatial separation [4]. Still, as a matter-of-fact, taking hours is justified and acceptable to the Geometric Modelling community. However, any design process is essentially a feedback cycle, and thus, is at stake due to such a long feedback time. After observing this for a while, I realized that a quick feedback is a need of the hour. Here was a rather deadlock situation: the complexity of the problem was not directly reducible, and at the same time, the speed of the machine was not under my (or anybody s) control. Their meeting point (10-11 hours) was not acceptable to the designer. After realizing this deadlock situation, the focus of thinking changed. I realized that a quick visual feedback is needed for the cycle. The geometric Modelling solution can still take its own sweet time of hours, once the designer has done CSI Communications July

26 all the necessary modifications using quick visual feedbacks in the cycle. The Solution The problem is now relatively easy: Create an image of the lump of 80 million triangles from one point of view quickly and show it to the designer. If it can be done in a minute or so, nothing is like that. If required, the designer may create one or two more views from different angles to see how it looks. Here the focus changed from Geometric Modelling to Image Synthesis. I started looking for Image Rendering techniques that handle very large data of scenes. Computer Games Design is a typical application dealing with such requirements. The tool I found to create the required image is called the z-buffer [5]. It accepts model data in the form of triangles, and applies high-end imagery techniques of ray tracing [6], scan-conversion [4], and occlusion culling [7] to them. The calculations are done by hard-wired implementation of the algorithms. The resulting image is directly accumulated in a huge fast-access memory buffer. The main advantage of using hardware implementation of z-buffer is speed. Another advantage is that it accepts any kind of collection of triangles without demanding any particular structural relation on them. The z-buffer typically has the capacity to handle one triangle in an average time of one millisecond. So for our 80 million triangles, it takes just 80 seconds slightly more than a minute. A designer may want to see from multiple directions, to get a better idea about how it looks. For more views, say three from x, y, and z directions, it takes 4 minutes. This gives a reasonable timing for the designer s feedback cycle. Addendum I did not stop at just that. The original problem of creating envelope was still there taking hours, even though the feedback cycle was fast. What I observed is that the envelope contains roughly the same number of triangles (30,000) like a single tire in canonical form. This is no surprise. It any way is just a surface, although it encloses a volume containing very large number (80,000,000) of triangles. The z-buffer was giving only one view of the envelope seen from some particular point and direction, and not the complete geometric model from all sides. However, at the same time this view was taking just 80 seconds. The solution now is to generate six views (or more if required) from six sides (say top, bottom, left, right, back, and front) Each view has roughly 30,000 triangles (total 180, 000). These of course overlap. One needs to develop appropriate software to (1) detect and remove repeated triangles from the collection, and (2) to stitch the six vies together so that the topology of the envelop is developed and its complete Geometric Model (as a Triangular Mesh) is established [1]. Such software for stitching was developed. It was taking roughly 8-10 minutes to do the stitching. This is 10 times more than the time required for rendering one view, but that is not exorbitantly large beyond expectations. The time mainly reduced because (1) the number of triangles have reduced from 80 million to 180,000, (2) a lot of coherence and spatial relations (structure) exists [4], and (3) more importantly, the time complexity [2] of stitching operation is not O(n 2 ) but between O(n) and O(n log n). This surely was not that bad from hours to 8-10 minutes! All this was achieved by changing focus and then applying techniques from one discipline (Image Synthesis) to solve problem in other discipline (Geometric Modelling). Acknowledgements The author is thankful to Dr. S. P. Mudur for many fruitful technical discussions and inputs to the topics discussed in this article. References [1] Hoppe, H: Progressive Meshes, SIGGRAPH 96, p99-108, [2] Aho/Hopcroft/Ullman: The Design and Analysis of Computer Algorithms, Addison-Wesley, [3] Mortensen M.E.: Geometric Modelling, Wiley & Sons, [4] Newmann/Sproull: Principles of Interactive Computer Graphics, McGraw- Hill, [5] Wand, M. et.al.: The Randomized z-buffer Algorithm: Interactive Rendering of Highly Complex Scenes, papers/siggraph01.pdf [6] Ellis et.al. : The Ray Casting Engine and Ray Representations, Int. J. Comput. Geometry Appl.'91,, [7] Zhang, Hansong: Effective Occlusion Culling for Interactive Display of Arbitrary Models, Ph.D. Thesis, Dept. of Computer Science, University of North Carolina at Chapel Hill, n About the Author Dr. Koparkar has a Ph.D. in Computer Science, in Since then he has published over 20 Research Papers in the prestigious International Journals and Conferences, mainly in the areas of Geometric Modelling, Image Synthesis, and Geometric Shape Processing in 2-D and 3-D. He has been on the International Journal Editorial Board and International Conference Program Committee. He has visited several organizations in different countries for delivering lectures, developing software and presenting research papers. He has been on various Academic Advisory Committees at the University and Government levels in India. He had worked in Research Institutes like TIFR and NCST, and in Corporations like Citicorp, Computer Vision, ADAC Laboratories (USA), and 3-dPLM/GSSL (India). He has written four Books: Unix for You, Pascal for You, Java for You, and C-DAC Entrance Guide. At present, he offers consultancy to corporate clients about various latest technologies CSI Communications July

27 Article Hareesh N Nampoothiri Research Scholar, University of Kerala, Thiruvananthapuram Distance Units in CSS3 CSS3 is the new standard for cascading style sheets recommended by W3C. CSS3 comes with a lot of exciting features which enable the designers to extend the possibilities of web design to further levels. Websites today are delivered across a variety of devices which supports all possible resolutions. The newly introduced distance units in CSS3 are a blessing for the designers which allows them to design adaptive websites without much effort. Distance units in CSS are used to define the lengths (in the form of width, height, size, thickness etc.) of various elements. In CSS, the length value is defined by a number (with or without a decimal point) followed by a unit identifier such as px, em etc. Two types of length units are used in CSS; absolute lengths and relative lengths. The absolute length units are fixed in relation to each other and are anchored to some physical measurement. These units are widely used in websites using fixed width layouts. Units such as mm, cm, in, pt, pc and px fall under this category. Unit mm cm in px pt pc Definition millimeters centimeters inches (1in = 2.54cm) pixels; (1px = 1/96th of 1in) points; (1pt = 1/72nd of 1 in) picas; (1pc = 12pt) Relative length units specify a length relative to another length property. Style sheets that use relative units can more easily scale for different resolutions and they are often used in variable width (fluid) layouts. Until CSS3, only the first two relative length units (em and ex) were available. The last five relative length units (ch, rem, vw, vh, and vmin) are the new units that are included in CSS3. The units ch Unit em ex rem ch vw vh vmin Definition Relative to the font size of the element. Relative to the x-height of the element's font. Relative to the font size of the root element. Relative to the width of the "0" glyph in the element's font. Relative to the viewport's width. Relative to the viewport's height. Fig. 1: HTML page showing the difference between em and rem and rem are font relative length units, where as vw, vh, and vmin are viewport relative length units. It is also possible to define lengths in percentage values. Relative to the minimum of the viewport's height and width. Percentage values are always related to another value, a length for example. In CSS recommendations, percentage values are considered as a number type. Hence, we are not including it in the list of relative length units. Font relative length units The rem unit The rem unit can be related to the em unit which is already available in CSS2. The em unit is equal to the computed value of the font-size property of the element in which it is used. The rem uses the computed value of the font-size of the root element. It can be defined in the style CSI Communications July

28 Fig. 2: The page coded in Fig. 1 as rendered in a browser Fig. 4: The page coded in Fig. 3 as rendered in a browser definition for the html tag or using the :root selector. Fig. 1 shows a sample HTML code which shows the difference in usage of em and rem. The font-size of the root element is defined as 60px. The width and height of the <div> elements is defined by two classes.applyem and.applyrem. The former sets the width and height of the <div> element to 1em and the latter sets these properties to 1rem. The width and height of both the <div> elements outside the <article> element is the same as both takes the computed value of the font-size property of the root element. But when both are applied inside <article> element it produces a different result. The font-size property of the <article> element is defined to 30px. The width and height of the <div> Fig. 3: HTML page showing the difference between ex and ch element which takes the.applyem class definition will be 30px. But the width and height of the <div> element which takes the.applyrem class definition remains the same, ie. 60px. Thus the advantage of using the rem instead of em is that, when we use nested elements, the value of the length units do not change unintentionally. The ch unit The ch unit can be related to the ex unit which is also available in CSS2. The ex unit takes the value of the height of character 'x' of the current font. The ch unit takes the value of the width of character '0' (zero) of the current font. In the example given in Fig. 3 we use Impact as the page font. The.applysqr class defines the height and width of the <div> element as 1ch and 1ex respectively. Thus the height of the <div> section will be the height of the 'x' character and the width will be the width of the '0' character of the font. The browser output is given in Fig. 4. CSI Communications July

29 Fig. 5: HTML page showing the usage of vw, vh, and vm Viewport relative length units The root viewport size is the width and height of the viewable area within the browser where the page is displayed. The units vw, vh, and vmin come under this category. All the three units are similar in nature. The vw unit stands for the viewport width and 1vw is equal to 1% of the current viewport size. Similarly, vh represents the viewport height. The vmin unit takes the value of vw or vh, whichever is smaller. When the height or width of the viewport is changed (by changing the browser screen size), the elements using viewport relative length units are scaled accordingly. (Fig. 6) Even though in W3C working draft the minimum value unit is given as vmin, we may use vm instead, as the modern browsers support vm instead of vmin. Note: The new length units included in CSS3 working draft is not supported by all browsers. The rem unit is supported by latest versions of Webkit, Gecko, Trident, and Presto based browsers whereas ch is supported by only Gecko and Trident based browsers. The vw, vh, and vm are only supported in latest versions of Trident based browsers. Recent versions of Webkit based browsers support vw and vh but not vm. References [1] CSS Values and Units Module Level 3 (dated 08 March 2012). W3C Working Draft (CSS3). Retrieved 2012 June 26, from [ w3.org/tr/css3-values/#lengths]. [2] Syntax and basic data types (n.d). W3C Recommendation (CSS2). Retrieved 2012 June 26, from [ html#values]. n Fig. 6: The page coded in Fig. 5, opened in two browser windows of different sizes About the Author Hareesh N Nampoothiri is a visual design consultant with an experience of more than a decade and worked with government organizations like C-DIT, C-DAC, University of Kerala and other private organizations. Currently, he is doing interdisciplinary research in ethnic elements in visual design in computer media. He is an author of two books on graphic design and a regular columnist in leading technology magazines including CSI Communications. Kathakli, blogging, and photography are his passions. He has directed a documentary feature on Kathakali and also directed an educational video production for IGNOU, New Delhi. CSI Communications July

30 Practitioner Workbench Baisa L Gunjal* and Dr. Suresh N Mali** * Assistant professor, Amrutvahini College of Engineering Sangamner, A nagar, MS ** Principal, Singhgad Institute of Technology and Science, Narhe, Pune, India Programming.Tips()» Multidimensional Plots in Matlab for Data Analysis Research outcomes and many application software outcomes need the analysis of data before its summarization. Using Matlab, experimental data analysis is done either using two-dimensional plots or three-dimensional plots. Program Listing 1: Two-dimensional Plots close all; clear all; t= 2002:1:2011 a= [9,9.5,9.7,10,10.2,10.5,11,11.5,12,12.5]; % Male Population b=[10,10.2,10.3,11,11.2,11.4,12,12.5,13,13.5]; % Female Population c=[8,8.5,9,9.2,9.5,10,10.3,10.5,11,11.5]; % Children population plot(t,a,t,b,t,c); h =plot(t,a,t,b,t,c); set(h(1),'markerfacecolor','red','marker','square') set(h(2),'markerfacecolor','green','marker','square'); set(h(3),'markerfacecolor','blue','marker','square'); h = legend('1:male Population ','2:Female Population', '3:Children population',4); grid on; xlabel ('Yearwise Population'); ylabel ('Population in Lacs'); title ('Population Plotting of Village for last 10 years'); and blue. We can assign an appropriate string to each line in the legend. We can also use the legend command to add a legend to a graph. The line width of plots can be adjusted as follows: set(h(3),'markerfacecolor','blue','marker','square','linewidth',1.5); Program Listing 2: Three-dimensional Plots [X,Y] = meshgrid(-10:.5:10); R = sqrt(x.^2 + Y.^2); Z = sin(r)./r; mesh(x,y,z,'edgecolor','blue') The output of this code will be: The output of above program will be: Fig. 2: Plotting three-dimensional plots using mesh Two- or three-dimensional bar charts are also used for representation of result analysis of given application. Matlab functions bar and barh draw vertical and horizontal bar charts. Two-dimensional vertical bars are shown in Fig. 3 a. Matlab function bar3 is used for drawing threedimensional bars as shown in Fig. 3 b. Fig. 1: Analysis of population using Matlab plots Legends' provide a key to the various data plotted on a graph. Fig. 1 shows the legend plotted with lines of different colors: red, green, Fig. 3: Data analysis a) Two-dimensional and b)three-dimensional bars About the Authors Baisa L Gunjal has completed MTech in IT and presently working as assistant professor in Amrutvahini College of Engineering Sangamner, A nagar, MS. She has 14 years teaching experience and she is the coordinator of seven postgraduate courses running in her college. She is also working on research project funded by BCUD, University of Pune and having more than 15 International publications including IEEE Explorer, IET-UK libraries etc. She is also a CSI Student branch coordinator at AVCOE Sangamner and CSI Member. Dr. Suresh N Mali has completed his PhD and working as principal in Singhgad Institute of Technology and Science, Narhe, Pune, India. He has written 3 technical books and published 25 papers in various national and international journals and various conferences. He is working as member of expert committee of AICTE and also worked as member of Local Inquiry Committee, University of Pune. He is member, Board of Studies for Computer Engineering in various universities like University of Pune, Shivaji University, Kolhapur etc. CSI Communications July

31 Practitioner Workbench Umesh P While using Python you may come across situations where you need to read data from a text file or write your results into a file etc. It is very easy to do this in Python. There are many ways of doing it. We can open a text file using Python and can read data from it or pass it to some variable. Otherwise the file can be opened outside Python using its default program. Depending on the context we can do in either way. A text file can be opened in Python by using its built-in functions - open() and readlines(). To read a text file, first we need to open the file by giving its path. f = open("path to text file.txt", r) Here the pointer to the text file is stored to the variable f. The second attribute in parentheses gives the path of the file that you are going to open. Here r stands for read mode of the file. There are other modes such as: w - overwrite to a file (if such file does not exist, then create and write) a - append to a file rb - read binary wb - overwrite a binary file r+ - open the file for both reading and writing a+ - open the file for both appending and reading. The following snippet illustrates how to create and write in a text file using Python text = open("file.txt", "w") l1= "This is my notepad file written using Python" text.writelines(l1) text.close() There is a simple way in Python to open a program. Let us see an example for opening an executable file outside Python. Write the following code and run in Python. import os os.startfile('c:\program Files\Google\Google Talk\googletalk.exe') This facility ensures seamless integration of Python programs with operating system utilities and applications. Excel files are the common data storing file format. We can read and write excel files using Python script. A package to support this task (xlrd package) needs to be installed. You can download the package and install or use easy installer for windows and Linux users can use the terminal for this (sudo apt-get install python- xlrd) If you are done with the xlrd package, try the following: import xlrd wb = xlsxrd.open_workbook('data.xls') For this, make sure that the excel file and Python script are in the same folder otherwise specify the path of the file. To work with xlsx files, you need to install and import xlsxrd. Try the following program, which reads data in an excel file and prints the data. fromxlrd import * book = open_workbook('data.xls') sheet = book.sheet_by_index(0) print sheet.name printsheet.nrows printsheet.ncols Department of Computational Biology and Bioinformatics, University of Kerala Programming.Learn ( Python )» Read and Write Using Python forrow_index in range(sheet.nrows): forcol_index in range(sheet.ncols): printcellname(row_index,col_index),'-', printsheet.cell(row_index,col_index).value We can select sheets by index. If there is more than one sheet in an excel file, we can choose a specific sheet and read file from that. Here in the second line, by book.sheet_by_index(0), the program selects the first sheet of the book. sheet.name, sheet.nrows, sheet.ncols prints the name of sheet, number of rows, and number of columns respectively. (row_ index,col_index) indicate the position of each element and(row_ index,col_index).value indicate the value in the position. Now let us write an excel file. For this xlwt module has to be installed. This can be done by following the instructions that has been given above. Then follow the codes given below. import xlwt book = xlwt.workbook(encoding="utf-8") sheet1 = book.add_sheet("new Sheet 1") sheet2 = book.add_sheet("new Sheet 2") sheet1.write(0, 0, "I have written first Cell of the First Sheet") sheet1.write(1, 0, "I have written in the second Cell of the First Sheet") sheet2.write(0, 0, "I have written in the First Cell of the Second Sheet") sheet2.write(1, 10, "This is written to eleventh Cell of the Second Sheet") book.save("new_excel.xls") Here book.add_sheet is the syntax for adding new sheet. sheet1. write followed by the cell id and text writes the text in the specified cell. Now try the following code to color from xlwt import * row = easyxf('pattern: pattern solid, fore_ colour blue') book = Workbook() sheet = book.add_sheet('precedence') fori in range(0,10,2): sheet.row(i).set_style(row) book.save('format10.xls') n CSI Communications July

32 Security Corner Adv. Prashant Mali [BSc (Physics), MSc (Comp Science), LLB] Cyber Law Expert Information Security» Privacy & Responsibility As a Panelist I am attending the India Security Meet 2012 to be held on 29th July in New Delhi. Thus, I thought of penning about the most talked topic of privacy. As the use of mobile phones and social networking, as well as resulting attacks, proliferate users personal information is subjected to more and more risk. Adding to that risk is the increasing connection between the cyber and physical worlds. As the college-going users of Facebook and Twitter grow older, they will still want the social networking capabilities they are used to, but will also be more concerned about privacy. There are companies out there developing and testing both secure social networking sites and privacy technologies to run on top of existing sites. For now, users should pay close attention to the kind of information the applications they use are sharing about them with others. Permissions granted knowingly or unknowingly. When speaking of physical systems, it is easy to recognize privacy concerns within the BFSI Segment as malware captures client information and sends it out across the Internet, or as busy managers carry vital data around on USB drives, DVDs, or on attachments. However, it is important to note that privacy concerns also come into play with the compromise of utility networks. A recent set of smart grid cyber security guidelines published by the National Institute of Standards and Technology (NIST) includes an evaluation of privacy issues at residences based on new smart grid technologies. According to IEEE Security & Privacy, Electricity use patterns could lead to disclosure of not only how much energy customers use but also when they re at home, at work, or traveling. When at home, it might even be possible to deduce information about specific activities (e.g. sleeping versus watching television). The data DTH users reveal to DTH operators or the easily manageable and accessible data in Call Data Records present with Mobile Operators. So who is responsible for all of this? Who is responsible for protecting users privacy and for halting the compromise Electricity use patterns could lead to disclosure of not only how much energy customers use but also when they re at home, at work, or traveling of computers, mobile phones, and other devices? Who is responsible for stopping the spread of malware and preventing the damage it could cause to our nation s critical infrastructure? It is very clear that the cyber security problem cannot be solved by a single group of people. Users, government, technology vendors, and security researchers all have a role to play in this fight, but each group alone can only go so far. We can t make the users responsible. Within an enterprise, the CSO has to be aware of what the real threats are and be dictating policies for the employees. Users have to be extra vigilant. Because so much of the initial infection today is driven by carefully crafted social engineering, botnet operators are successful even against computers that have practically every protection technology known to man. Having said that, those layers of defense should not be neglected - at the very least they limit the scope of attack. The solution is not just technologybased or policy-based, but requires a more holistic approach to obtaining a deeper understanding of the threats through the collaboration of users, government, academia, and industry. The name of the game today is to know what you don t know. Staying plugged into external environments and the overall threat Vector is a key for being prepared for when attacks really do emerge. Today, security has gotten so complex that there is no way a single person can even know everything about one aspect of cyber defense. It is therefore critical for leaders in the security industry to share information with one another. Government and industry collaboration is the key when it comes to protecting physical systems from cyber attack. When you consider how much of our critical infrastructure is owned and operated by the private sector, it becomes clear that there is a need for greater public/private partnership when it comes to mitigating risk. Moving forward, government organizations that possess classified information about potential threats will need to regularly share this actionable intelligence with the private sector in a more timely and structured manner to effectively defend our nation against attacks. The threat is now so big that the old style of developing a separate remedy for every threat simply does not scale, so a community-based defense approach is the need of the hour. A cultural revolution towards Cyber Security and Privacy is must, and this can be brought out through larger cooperation and awareness amongst stake holders. Moving forward, government organizations that possess classified information about potential threats will need to regularly share this actionable intelligence with the private sector in a more timely and structured manner to effectively defend our nation against attacks. The threat related to loss of privacy in this cyber insecure village relates to all. It is therefore up to all of us to educate ourselves on the various cyber security risks and do our part to stop enabling and spreading malicious cyber activity affecting our privacy. n CSI Communications July

33 Security Corner Mr. Subramaniam Vutha Advocate IT Act 2000» Prof. IT Law in Conversation with Mr. IT Executive: Issue No. 4 IT Executive: Hi Prof. IT Law! The last time we met, you showed me how I often enter into electronic contracts without even realizing that I have done so especially on websites. Prof. IT Law: Yes, the law of contracts is broad enough to permit two parties to bind themselves to a contract orally, in writing, or even by actions that convey their intent. As for example, when you pick up a newspaper from a vendor and give him Rs. 5 for it. IT Executive: What happens in real world situations? Are the same contract principles applicable to web-based contracts? Prof. IT Law: Yes, the principles of contract formation are the same whether you get into contracts in the real world or in the virtual web-based world. But the web poses some special challenges to contract formation. IT Executive: For example? Prof. IT Law: You remember that I told you that all it takes to form a contract in real life is for one person to make an offer and for another to accept it? IT Executive: Yes I do. Prof. IT Law: So let us consider what happens when you enter a shop. The shop has many items on display, and each item has a price tag on it. When you pick up an item and pay for it you have a classic case of an offer and an acceptance of that offer made orally. Often without even a word being spoken! IT Executive: I see. But in this case who makes the offer and who accepts the offer? Prof. IT Law: Good question. And the answer is crucial for web-based contracts too. The law says that when you enter a shop and see items with price tags, it is you who makes the offer to the shopkeeper when you pick up an item and offer cash. IT Executive: And the shopkeeper accepts my offer by accepting my cash? Prof. IT Law: Yes and a contract is formed. It is as simple as that. But now consider what happens if the shopkeeper refuses to sell you the item. For example, by refusing to take the cash you offered, or your credit card. IT Executive: I guess he can do that because I am making an offer to buy the chosen item, and he can either accept or refuse the offer. Prof. IT Law: You are right. And so is the reason given by you. If it was the other way around, i.e. if the shopkeeper makes an offer to you by displaying items with price tags that you could accept then, in that case, the shopkeeper would not be in a position to refuse to sell you the item. And the shopkeeper would be bound by that contract. IT Executive: I see. But how is this crucial in web-based contracts? Prof. IT Law: As in shops, the question of who makes the offer is crucial even on websites. For example, if you make an offer on your website and it is accepted by a visitor to your site, you would have a contract that binds you. IT Executive: That is ok I guess. Prof. IT Law: Not always. You may not want to ship products to someone who accepts your website-based offer because he lives in another state or country. Or you may have run out of stock. Or you may have found it necessary to revise your prices, and so on. IT Executive: So what is the solution? Prof. IT Law: Here it is. Your website is like a shop. And you display goods on it with price tags. But the offer is made by the visitor to your site. That is what legal scholars say. So you can refuse the offer made by the visitor to your site. Just like the shopkeeper. IT Executive: That is good for those who sell goods on websites. Prof. IT Law: Yes, the principle is the same whether it is a shop or a website. But you must be careful not to say anything on your website that reverses these principles. For example, if your website screams in large bold fonts that your offer closes at 5 p.m. on 30th June, 2012, it may be difficult for you to say later that you were simply inviting offers from visitors to your site. IT Executive: Can you tell me more? Prof. IT Law: To make sure that your website makes invitations to people to make offers to you for the items you are selling, you will need to design your website accordingly. For example, if your website repeatedly uses the words OFFER, OFFER, OFFER or OFFER closes by 5 p.m. on Sunday, 24th June, 2012, it may be difficult to then contend that you were not making an offer that visitors to your site can accept. IT Executive: And how about an I agree button rather than an I accept button for those who wish to buy the items displayed? Prof. IT Law: Yes, that is a good idea too. In sum, you should have your lawyer apply his or her mind to the design of your website to avoid problems with transactions on your site. IT Executive: Ah, that is good to know. Thank you. I look forward to meeting you again soon. Prof. IT Law: Yes, I enjoy the sessions with you too. See you soon. n CSI Communications July

34 Society Achuthsankar S Nair Editor, CSI Communications Upload India, Upload The free and open source software movement is here to stay. It has become a strong alternative to proprietary software. What is more, the philosophy professed by the movement has been imbibed by many a field of intellectual activity such as literary publishing, music, photography etc. That knowledge should be freely available is beyond debate for a beneficiary. However, the act of deploying free the knowledge generated or held by oneself, is another cup of tea. The enthusiasm for the former is not seen in the same measure for the later. India was once a knowledge powerhouse (along with Persia, China, and Greece). Today, we are in the process of re-emerging as a knowledge power. But if we make a casual comparison of our cyber landscape with that of the west, we may see a chasm. If you ask an average Internet user working on a computer whether s/he is uploading or downloading, the answer is anyone s guess. Except for photographs or posts that are sprinkled up on the social networking sites, and some impressive activity in blogging and wikying, we are far from the center of gravity of the cyber knowledge world. Take for instance our Universities (the thousands of colleges are of course even worse). Many of them have utility websites that serve the purpose of their governance. They process online applications, announce results, some accept fees online, and most give away administrative information and also a touch of history and current news. Only a few have institutional repositories, and even lesser have teaching material including lecturing videos that go up on the web. Contrast this with MIT that has an open courseware initiative, which throws up online all lectures live along with the associated documentation like handouts, power-points etc. Some public institutions hold precious collections of manuscripts, books, and other documents, which have been digitized but not available online. I reiterate the self-criticism about our adoption of freesoftware philosophy - we are good at taking knowledge free, but not so good at giving it free! I do not want to be seen as a pure skeptic. I see silver linings. Initiatives by state University like Mahathma Gandhi University to create a searchable archive of all PhD Theses (mgutheses.org) and a great initiative from IISc in cooperation with CMU, IIIT, NSF, ERNET, and MCIT and 21 participating centers to digitize significant works of mankind into a Digital Library of India and also the eprints@iisc repository, which collects, preserves, and disseminates in digital format the research output created by the IISc research community, are all initiatives that keep our hopes alive. To be equal citizens in the cyber-world, we need more initiatives like these, so that we do not become a nation of mere knowledge exploiters. It also will be a great shame in view of the tremendous potential that this country s human resources represent. Upload India, Upload. n CSI Communications July

35 Brain Teaser Dr. Debasish Jana Editor, CSI Communications Crossword» Test your Knowledge on Image Processing Solution to the crossword with name of first all correct solution provider(s) will appear in the next issue. Send your answers to CSI Communications at address with subject: Crossword Solution - CSIC July CLUES ACROSS 1. Type of ratio representing the width of an image divided by its height (6) 4. Tool named matrix laboratory (6) 6. A range of gray shades from white to black (9) 9. A color model representation (3) 10. A morphological operation on image (8) 11. A standard vector image format (2) 13. A method to compute a new value of central pixel in a neighborhood (11) 15. One primary color (5) 16. An image processing and GIS software package (5) 18. Type of matrix that remains unchanged in value following multiplication by itself (10) 23. Difference between the lightest and darkest regions of an image (8) 26. Color in the HSB color model (3) 27. The ability to distinguish fine spatial detail (10) 29. A geometric image transformation process (8) 31. Type of image transformation that preserves straight lines (7) 32. Binary digits (4) 33. An image reproduction technique where the various tones of gray or color are produced by ink dots (8) DOWN 2. A device used to take pictures (6) 3. Process that makes each pixel more like its neighbors (8) 5. A geometric process that reduces image size (8) 7. A dark area or shape produced by a body coming between rays of light and a surface (6) 8. Microsoft format for audio and video files (3) 12. An 8-bit-per-pixel bitmap image format (3) 14. Type of filter that attenuates the high frequency information in an image (7) 17. The process of recording an analog signal in a digital form (12) 19. An image display technique (9) 20. The science of measuring human color perception (11) 21. A graphical display of tabulated frequencies (9) 22. An imaging sensor (3) 24. The physical measure of brightness (9) 25. A geometric image transformation process (8) 28. Representation of picture element (5) 30. An image file format (4) Do you know Lena? Lenna or Lena is one of the most widely used figures in image processing research. The name of the lady is Lena Söderberg, a Swedish model, born 31st March, 1951, shot by photographer Dwight Hooker for Playboy (original appearance in Playboy centerfold in Nov 1972). She was invited as a guest at the 50th Annual Conference of the Society for Imaging Science and Technology (IS&T) in (a) Picture of Lenna taken in May 1997 at the IS&T's conference (Ref: edu/~chuck/lennapg/lenna.shtml) (b) Widely used picture of Lenna, image source: Signal and Image Processing Institute at University of Southern California (USC- SIPI) image database (Ref: Congratulations to Mrs. P Deepa (Panimalar Engineering College, Chennai) and Dr. Vasudeva Acharya (Srinivas Institute of Technology, Valachil, Mangalore) for getting ALL correct answers to June month s crossword. Solution to June 2012 crossword 1 S E M A N T I 2 C S 3 T C 4 D A T A M I N I N G H X 5 K 6 N O W L E D G E O E M 7 A N A L 8 Y S I S 9 T U 10 T A G O E 11 H E U R I S T I C 12 X M L M R A Y A M 13 L E A R N I N G 14 M 15 E N 16 R N E X 17 I T 18 C O N C E P T 19 T A C I T P N I L T A E F C A 20 W I S D 21 O M 22 B K R O W 23 I P R O E O N T R E S 24 C R M T O O S 25 M D B P K A K W Y A 26 D S S D M L S T 27 C A A E 28 T R A I N I N G 29 S T R U C T U R E D E O B A K G M 30 K N O W L E D G E B A S E E CSI Communications July

36 Ask an Expert Dr. Debasish Jana Editor, CSI Communications Your Question, Our Answer A computer would deserve to be called intelligent if it could deceive a human into believing that it was human. ~ Alan Turing Subject: PHP arrays What are the different types of arrays in PHP and provide few examples for illustration of their use. Thanks. Anonymous A In general, an array can store one or more values and referred in a single variable name. Each element in the array can be assigned its own identifier (we call index or key) so that the corresponding element stored can be easily accessed through the key, for example: $array[key] = value; In PHP, arrays can be three types: Numeric Array Associative Array Multidimensional Array A numeric array stores each element associated with a numeric key value (like array index in C/C++/Java). We can define the numeric array like below: (a) $anames = array("amar","akbar","anthony"); Or (b) $anames[0] = "Amar"; $anames[1] = "Akbar"; $anames[2] = "Anthony"; The array and array elements can be used in a PHP script, for example: <?php $anames[0] = "Amar"; $anames[1] = "Akbar"; $anames[2] = "Anthony"; echo $anames[1]. " and ". $anames[2]. " are ". $anames[0]. "'s friends";?> The output would be displayed as: Akbar and Anthony are Amar's friends In an associative array, each key is associated with a value, could be numeric or nonnumeric. We can use the some values as keys and assign some other values to them. Using an associative array we can assign number of chocolates belonging to a person now: $chocolates = array( Amar"=>4, Akbar"=>5, "Anthony =>6); Or, $chocolates [ Amar '] = 4"; $chocolates [ Akbar '] = 5"; $chocolates ['Anthony'] = 6"; The associative array and array elements can be used in a PHP script, for example: <?php $chocolates [ Amar '] = 4"; $chocolates [ Akbar '] = 5"; $chocolates ['Anthony'] = 6"; echo Akbar has. $chocolates [ Akbar ]. "chocolates.";?> The output would be displayed as: Akbar has 5 chocolates. In a multidimensional array, array may contain another array and can be nested to any level. Each element in the main array can also be an array. And each element in the sub-array can be an array, and so on. For example: $friends = array ( "FA"=>array ( "Amar", "Akbar", "Anthony" ), "FB"=>array ( "Raja" ) ); echo $friends ['FA'][1] The output would be: Akbar Accessing Arrays can be done in many ways. One may use direct access to obtain the item. For sequential access, the foreach loop was designed to work with arrays. This iterates through the items in two different ways: foreach ($array_element as $key => $value) It would provide both the key and value at each iteration foreach ($arrayvar as $value) Provides just the next value at each iteration. Ex: Using each to iterate: The each function returns a pair with each call, (a) a key field for the current key and (b) a value field for the current value. It returns the next (key,value) pair, then moves forward while ($x = each($array)): $key = $x ["key"]; $val = $x["value"]; echo "key is $key and value is $val<br>\n"; endwhile; n Send your questions to CSI Communications with subject line Ask an Expert at address csic@csi-india.org CSI Communications July

37 H R Mohan AVP (Systems), The Hindu, Chennai hrmohan.csi@gmail.com ICT News Briefs in June 2012 The following are the ICT news and headlines of interest in June They have been compiled from various news & Internet sources including the financial dailies - The Hindu, Business Line, Economic Times. Voices & Views London 2012 Olympics will be biggest smartgame ever'. There will be real-time information and results to the world's broadcast media in less than 0.3 seconds to broadcasters globally. The cloud computing market is expected to grow to $241 billion in 2020, against $40.7 in Mr. B Suresh Babu, Additional Commissioner at Commissionerate of Industries, AP. As many as 604 universities and 35,000 colleges will be brought under the national knowledge network (NKN) in the next six months - Mr. Kapil Sibal. Apple accounted for 11.8 million (68%) of the 17.4 million tablets that were shipped in IDC. Income from financial services, telecom segments of IT firms takes a beating. Airtel, Idea, Vodafone face Rs 1-lakh crore outgo on re-farming. 50% car buyers do online research - study by Neilsen. IT spend in financial services to touch Rs. 38K crore - Gartner. IT, ITES firms in AP clock Rs. 50,000-crore biz. Nasscom maintains growth forecast of 11-14% for IT industry. Cyber security must be top priority for CEOs - Deloitte. Over 85% enterprises to deploy tablets this year - Gartner. Smartphones seeing malware explosion - Software security companies. India among worst affected by malware. Malware detected on 13.8 of every 1,000 computers scanned in India as against the global average of Microsoft. Top PC makers logging out of netbooks as sales drop. Global R&D spending rises 8.2% in FY11 - Zinnov. Phishing attacks on.in domain rising - Symantec. In 2011, 7% of consumer content was stored in cloud, but this will grow to 36% in Gartner. Worldwide consumer digital storage needs will grow from 329 exabytes in 2011 to 4.1 zettabytes in Average storage per household to grow from 464 GB in 2011 to 3.3 TB in Gartner. Over half (53%) of children in India have been bullied online - Global Youth Online Behavior Survey by Microsoft. Enterprise social software market size is growing and will touch $6.4 billion in IDC. IT sector feels the Obama heat and slows in Telecom, Govt., Policy, Compliance Spectrum: GoM to decide on TRAI, DoT panel pricing formula. Airtel, Hughes unable to zero-in on users, says Intelligence Bureau. 90% cable TVs yet to go digital in Kolkata while it is hardly one month to go for the digitization deadline. JPC on 2G to submit report in six months. Govt. to remove multilevel TDS on software from 1st July, DoT to finalize auctioneer for spectrum auction by July 10. Specifications for Aakash 2 revised version to be finalized by June. TRAI has hauled up mobile companies for failing to comply with its order on seeking consumer's consent before activating VAS. Hardware makers ask Govt. for cover against forex volatility. 2G Scam: JPC may decide against summoning politicians. TN Govt. wants BPOs to open units in village panchayats. Capital subsidy of 20% to be provided. India ready to move to new Internet protocol IPv6. Satyam case: 44 properties of promoter's family attached. Raja asserts his innocence on 2G charge. Bharti Airtel to pay Rs. 700 crore for customs duty evasion case. AP to extend e-governance initiative Mee Seva to all districts. All panchayats to have fiber optic connectivity in three years - Sachin Pilot. IT Manpower, Staffing & Top Moves IT Industry witness 29% decline in hiring - Mr. Rajesh Kumar, CEO, MyHiringClub. com Fashion and You to hire 400 people by this fiscal end. Only 1 out of 7 opts for Nokia Siemens' voluntary retirement scheme at Kolkata unit. PayPal zooms in on IITs for hi-tech' hires. Has hired 80 graduates at Rs. 7 lakh a year. Forbes describes Mr. Sridhar Vembu, founder and CEO of Zoho, as the Smartest Unknown Indian Entrepreneur'. Zensar to hire 1200 engineers in Karnataka this fiscal. Over 17,000 foreign nationals on rolls in TCS. Majority of Indian employees approve personal use of social media at work. USCIS reaches cap on H-1B visas. Nokia to cut 10,000 jobs. Subex grants ESOPs to employees. Microsoft launches Wi-Fi bus service for employees. IT firms have made over 1 lakh job offers so far this year - Nasscom President. Bangalore IT firms caught by hunger pangs due to a curtailed 11 p.m. dinner time. Som Mittal to continue as President of Nasccom for another two years. RIM to cut about 5,000 jobs. Company News: Tie-ups, Joint Ventures, New Initiatives Google tips China searchers to hotbutton terms that evidently prompt censors to derail queries. Microsoft releases final test version of Windows 8. Intel unveils new chip for faster computing devices. Dell introduces laptop battery recycling program. Infosys has been identified as one of the top 25 performers in Caring for Climate Initiative by the UN Global Compact and UN Environment Program. Mahindra Satyam, Tech Mahindra shareholders okay merger. Allied Computers to launch laptops at Rs. 4,999. Agiliq Info Solutions announces a tool that converts the blog into an app. Wipro AppLife Contest for engineering students open from June 19, 2012 to December 18, Microsoft unveils Surface tablet running Windows RT and Windows 8 Pro. Bharti Airtel launches Behtar Zindagi, a mobile-based service for farming. Duke University, UK, develops gigapixel camera. Printer virus hits BPO, health care, manufacturing cos. HP to pay Rs lakh for selling defective laptop. Intel launches future scientist program in Karnataka. To train 500 teachers and 5,000 students. Microsoft offers Office 365 at no cost to educational institutes. TCS(China) Co. Ltd., has been recognized as 2012 Top 10 Global Service Providers in China. Google unveils Nexus 7 tablet designed to challenge Apple s ipad. Beware of offers this Olympics season! - Mr Amit Nath, Trend Micro. n CSI Communications July

38 CSI Report Dr. Dharm Singh Convener SIG-WNs CSI WTISD-2012: World Telecommunication and Information Society Day (L to R: Dr. Deepak Sharma, Smt. Ridhima Khamesra, Er. Sushil Kabra, Prof. R K Aeron, Dr. Y C Bhatt, and Dr. Dharm Singh) The World Telecommunication Day was celebrated on 17th May, 2012 by the Institution of Engineers (India), Udaipur Local Centre in association with SIG-WNs & e-agriculture CSI at IEI Udaipur Local Centre. This year the theme of the World Telecommunication and Information Society Day was Women and Girls in ICT. Story Behind the Celebration: World Telecommunication and Information Society Day, celebrated each year on 17th of May, marks the anniversary of the signature of the first International Telegraph Convention in 1865, which led to the creation of the International Telecommunication Union. Girls in ICT Day: In order to emphasize the theme of WTISD-12, ITU also marked international "Girls in ICT Day" this year on 26th April Girls in ICT Day, to be held every year on the fourth Thursday of April, encourages members to host events where girls and young women are invited to ICT companies and government agencies to appreciate the opportunities the ICT sector holds for their future. Women are the bedrock of our societies. They are the pillars of strength in every family and community. Yet gender inequalities remain deeply entrenched. Women and girls are denied access to basic health care and education and to equal opportunities at work. They face segregation in economic, political, and social decision making and often suffer violence and discrimination. ICTs are tools: ICTs are tools that can help accelerate progress towards achieving this target, and ICTs related e-applications are key instruments providing basic services and achieving the Millennium Development Goals, such as providing community health care, safe drinking water and sanitation, education, food and shelter; improving maternal health and reducing child mortality; empowering women, girls and the more vulnerable members of society; and ensuring environmental sustainability. Chief Guest of the function was Prof. R K Aeron, former chairman, CSI Udaipur Chapter. He addressed that computer and knowledge of ICT should be used as value addition to your work, and besides computer more stress needs to be given on fundamental learning of the subject concerned. The Guest Speaker Er. Sushil Kabra, DGM, BSNL, Udaipur highlighted the ICT Infrastructure and Challenges. He informed the house that 28,672 kiosks have been planned in India, and out of this 85 will be in Udaipur Region. Under NOFN work the third tier of Govt., i.e. Panchayats, will be focused for access of information on education health, financial, and systematic services in rural sector. This is a project to connect 2, 50,000 GPS with 100MBPS broadband connectivity at each GP within time frame of 2 years with funding of Rs. 20,000 crore executed by SPV called BBNL. The major challenges in expansion of ICT infrastructure are language, trained human resources, power, ROW permission & charges, and O&M issues. Audience during celebration of WTISD-2012 Another Guest Speaker Smt. Ridhima Khamesra, Head, Department of CSE, GITS, Udaipur, quoted the words of UN Secretary General that There is gender divide with woman and girls enjoying less access to ICT than men and boys in She informed that lakh woman employees were engaged in organized sector with 58% in public sector and 42% in private sector. The percentage of job seekers has increased to about 27.8% in Despite rapid growths of ICT jobs in past few years, less than 20% woman account for it which is less than 5% of total female population. She stressed that there can t be real knowledge economy if women are not active participants of the story. At the start of the program welcome address was given by Dr. Y C Bhatt, Convenor, SIG-e Agriculture, CSI and Chairman of the IEI Udaipur Local Centre. He emphasized that ICT plays a very important role in the modernization of rural services where a clear digital divide is visible in the country for education, health, and agriculture sector. The advancement of ICT in agriculture sector for precision technology/extension will make pathway for sustainable development of rural India. The program was conducted and at the end vote of thanks was given by Dr. Deepak Sharma, Committee Member, IEI, ULC. CSI Communications July

39 CSI News From CSI Chapters» Please check detailed news at: SPEAKER(S) LUCKNOW (REGION I) Dr. Bharat Bhasker, Mr. Arvind Kumar, Mr. Navjot Singh, and Mr. Pradeep Kumar TOPIC AND GIST 21 April 2012: Storage Technology Day-2012 Dr. Bhasker delivered speech on Facilitating High Performance Computing and Mr. Arvind Kumar delivered speech on Growing Career Opportunities with Changing Trends of Data Storage Technologies. Industry experts Mr. Navjot Singh and Mr. Pradeep Kumar, delivered talks on emerging trends in storage technology and appreciated efforts of Integral University for producing large number of storage professionals in India. KOLKATA (REGION II) Mr. Vinod Agarwal, Dr. J K Mandal, and Mr. Arijit Roy Guests on dias 19 May 2012: Workshop on Information Security and Enterprise Risk Management Mr. Vinod Agarwal spoke about importance of Information Security and Risk Management. Dr. J K Mandal introduced the participants to the key elements of information security as applicable to e-commerce Applications. Mr. Arijit Roy spoke about fundamentals of Enterprise Risk Management and the concept of Risk Intelligence. PUNE (REGION VI) Dr. Deepak Shikarpur and Mr. Amit Dangle Amit Dangle, Rahil Shah, Soumi Alphons, and Capt. Mahesh Jog During the lecture 22 May 2012: Meeting with Mayor and Commissioner of Pune Mayor of Pune, Mrs. Vaishali Bankar and Commissioner of Pune, Mr. Mahesh Pathak organized a meeting with community in Pune. This interaction and brainstorming led to various ideas to improve e-governance, Coordination in civic administration, improved Internet bandwidth, GIS-based solution to ease Road digging, Pune 2020 vision with 100% literacy, employability, IT managed public transportation/traffic Signal synchronization, Recycling of e-waste, Rain water harvesting, and state-of-the-art citizen-centric e-governance. Meeting with Mayor of Pune 7 June 2012: Walk for Health Pune Chapter organized Walk for Health" initiative with an objective of raising awareness about the health risks that professionals from the IT industry face. This event organized in Magarpatta also highlighted measures that can be implemented by these professionals to address health-related concerns. The event saw good participation from different IT firms, which come together to spread the awareness about healthy living. The "Walk for Health" event included a symbolic walk inside Magarpatta city. Before starting walk for health CSI Communications ions July

40 SPEAKER(S) COCHIN (REGION VII) Mr. Vijaykumar Nair K TOPIC AND GIST 4-5 May 2012: Two-day workshop on Android In association with IEEE CS this two-day workshop on Android Bootcamp Training was conducted. This was a hands-on training for designing and building mobile applications using Android open-source platform, which explained the philosophy of developing for Android through its main application development building blocks and their interaction with one another. Mr. Vijaykumar Nair taking a session during the Android Workshop Mr. Vijay S Paul 11 May 2012: Technical talk on Social Media Marketing This talk was organized in association with IEEE CS for celebrating the National Technology Day. The speaker talked about basics of why businesses need to focus on Social Media Marketing in the present day world, and the growth of "Relationship Marketing" over "Broadcast Marketing". It also gave an outline as to how each platform is being utilized by brands to interact with their consumers. Mr. Vijay S Paul taking a session on Social Media Marketing From Student Branches» SPEAKER(S) TOPIC AND GIST AMITY SCHOOL OF ENGINEERING & TECHNOLOGY, NOIDA (REGION-I) Prof. (Dr.) Balvinder Shukla and Prof. (Dr.) Abhay Bansal 30 May 2012: Workshop on FDP on Open Source Software Faculty Development Program (FDP) was divided into various sessions covering topics, such as basics of open source software, Drupal basics, installation intricacies, module usage and theme selection. The participants got hands-on practice on Drupal, a framework for content management system. Prof. (Dr.) Balwinder Shukla, Acting Vice Chancellor, AUUP &DG ASET addressing the participants ACROPOLIS INSTITUTE OF TECHNOLOGY & RESEARCH (AITR), INDORE (REGION-III) Mr. Sanjeev Agrawal, Dr. Kamal Bharani, Mr. Achal Jain, February 2012: 5th National Level Technical Festival TechFest' 12 Prof. Sanjay Bansal, and Mr. Anshul Kosarwal TechFest 12 involved a variety of events such as MATLAB Programming Contest, T-Shirt Painting, Utkarsh (Paper Presentation), Project Presentation, Chakravyuh (Programming Competition), Eco-Mansion, Robo-war, Roborace, Cognizance (IT-Quiz), Nemesis (LAN Game), Gully Cricket, Color Jam (Wall Painting) and Kshitij (GK Quiz). Other special events were Zorbing, PaintBallFight, Rope Climbing, Electronic Mayajaal (Circuit Design), Source Scene (Code Debugging), Aahvaan (Case Study), Xtra- Inning (Freaky Flips, Final Destination, Gaon ki Masti), Shelter Design for all, and Street Soccer. ARMAGEDDON 2012 POSTER CSI Communications ions July w.cs indi

41 SPEAKER(S) TOPIC AND GIST AES INSTITUTE OF COMPUTER STUDIES (AESICS), AHMEDABAD (REGION-III) Ms. Trushali Jambudi and Mr. Vinay Vachharajani 16 April 2012: Workshop on Web Page Designing Using HTML5 and CSS3 The workshop focused on the features of HTML5 related to web page designing like Audio and Video support in HTML5, Sectioning tags and Form tags as well as features of CSS3 such as Background and Color Gradients, Fonts and Text Styles and List styles and Table Layouts. The participants were given hands-on exposure for the features covered in the workshop. Ms. Hiral Vegda and Ms. Amita Jagirdar Prof. Trushali Jambudi, during the workshop 27 April 2012: Workshop on Working with C#.NET During the workshop, participants were introduced to one of the objectoriented programming languages viz. C#. The workshop covered basic concepts of C#, features of C#, Why C#, and Comparison of VB and C# language. The participants were trained to create new applications using the basic concepts of C# like conditional statements, loop control structures, classes, reference and out parameters and MS SQL Server database. They were given online demo of all the concepts discussed. Dr. Rammohan Speaker during the session C#.NET 4 May 2012: Expert Lecture on Research Areas in Artificial Intelligence The speaker spoke on a wide range of topics in research. He spoke as to which are different areas for research in Artificial Intelligence, also he spoke on Soft Computing and Natural Language. He explained what the relation between Artificial Intelligence & Soft Computing is. He discussed various examples on Soft Computing. He told how to start research in this subject. He informed which journals students should refer in this subject. ITM UNIVERSITY, GWALIOR (REGION-III) Mr. N S Choudhary and Prof. Anupam Shukla Mr. Chouhan Dr. Rammohan giving lecture on Research areas in Artificial Intelligence 3 March 2012: One-day workshop on Advanced Computing & Robotics Mr. Choudhary spoke about concept of advanced computing, which is used in super computers or in computer cluster to handle major projects. Advanced computing is also used in simulation, for modeling of stars in Astrophysics, and to know new protein structure by biologist. Advanced computer users are also called power users. Mr. Choudhary also spoke about turing machine and translation process of natural language. Prof. Anupam Shukla told how to use robotics in designing, and in production process in industry. He explained how robot works and finds its destination using algorithms. Guests on dias 3 May 2012: One-day workshop on 3D Animation & Film Making Initially, Mr. Chouhan discussed about Maya Software and working on the software. Later he explained in detail concepts such as Animation, Croma, 3D, Sound Effect, Pre-Production, Modeling, Texturing, Lighting, Rigging, Rendering and developed a live project of flying butterfly. Later, he showed some clips of films like Ra-One, Chak De India etc. and explained the use of animation. Speaker conducting the workshop CSI Communications ions July

42 SPEAKER(S) TOPIC AND GIST R.V. COLLEGE OF ENGINEERING, BANGALORE (REGION-V) Mr. Balaji V, Mr. Anand B S, Mr. Jagan Jothivel, 28 May 2012: One-day Seminar RVCE- CISCO Day Dr. B S Satyanarayana, Prof. K N Raja Rao, Dr. Satyanarayana told that network is the key to knowledge enhancement. Dr. N K Srinath, and Prof. Chandrashekar Prof. K N Raja Rao expressed that the facility of setting up of a Lab by Cisco needs to be effectively utilized by the students. Dr. N K Srinath briefed about the course scheduled by Cisco Network Academy (RVCE). Later other speakers from Cisco expressed that every professional must know about network concepts and should have practical exposure. The future demand for skilled workforce in network field is quite high. Dignitaries during the RVCE-Cisco Day SRINIVAS INSTITUTE OF TECHNOLOGY, MANGALORE (REGION-V) Dr. R K Shyamasundar, Mr. Shekhar H M P, Prof. K Chandrasekaran, Prof. B B Amberkar, and Prof. Manjaiah D H 19 April 2012: National conference on Recent Research Trends in Network Security The AICTE sponsored national conference was inaugurated by Dr. R K Shyamasundar. Various speakers gave special lectures on different topics of network security. Dr. R K Shyamasundar addressing, sitting Prof. Shashidhar Kini, Dr. Shrinivasa Mayya D, Sri. CA A Raghavendra Rao S.R.K.R. ENGINEERING COLLEGE, ANDHRA PRADESH (REGION-V) Dr. G P S Varma 21 March 2012: Making a rakhi of 30 feet diameter and 300 feet length Final year Students of department of Information Technology, made an attempt to break the limca book of records by making a rakhi of 30 feet diameter and thread of length 300 feet. Students felt that study is not the only milestone that one has to pass; hence in order to show responsibility for environment and society, they worked out on this record. The inspiration behind the initiation of this activity was Dr. G P S Varma and senior students, who made a paper bag of large size and recorded in the limca book of records last year. DR. D.Y. PATIL INSTITUTE OF MCA, AKURDI, PUNE (REGION-VI) Mr. Amit Samadar and Mr. Amitabh Purohit 10 March 2012: Industry-Institute Interaction Programme Institute Interaction Programme was the platform for MCA students to interact with the industry experts. Mr. Amit Samadar guided students on various Industry expectations and spoke about challenges for freshers. Mr. Amitabh Purohit gave some guidelines and shared his experience on how to prepare for an interview (Group Discussion skills, Personal Interview skills, Technical interview skills etc.) with students to prepare them for their forth coming placement activities. He also conducted QA session on interview skills. Participants: SYMCA students (Div A & B) CSI Communications ions July w.cs indi

43 SPEAKER(S) TOPIC AND GIST MARATHWADA INSTITUTE OF TECHNOLOGY (MIT), AURANGABAD (REGION-VI) Dr. D G Regulwar, Dr. M S Joshi, Dr. S S Sane, Dr. V R Ratnaparakhe, Dr. A S Bhalchandra, Prof. R B Naik, and Prof. M B Nagori 30 April 5 May 2012: Program on "Research Methodologies and Advances in Computer Sciences" The main aim of this short-term training program was to make the participants conversant with Research opportunities in the field of Data Mining, Real Time Systems, Pattern Recognition and Fuzzy Logic. It also provided a platform to learn about Research Methodologies, Data Analysis and Optimization Techniques from experts of reputed institutions. During the session MAHARAJ VIJAYRAM GAJAPATHIRAJ (MVGR) COLLEGE OF ENGINEERING, VIJAYNAGARAM (REGION-VII) Mr. Arun Prathikantam March 2012: Workshop on Android and J2EE Mr. Arun Prathikantam gave introductory talk on Android and J2EE. He also demonstrated coding and simulation of the CRM app on android. The workshop was followed by hands-on exercise on J2EE concepts like STRUTS, WEBSE and latest technologies. Mr. Arun Prathikantam during workshop MAHENDRA ENGINEERING COLLEGE (MEC), NAMAKKAL (REGION-VII) Dr. Chee Peng Lim, Dr. V Prithviraj, Dr. P Naghabushan, and Dr. T Ravi March 2012: Three-day International Conference on Innovative Computing and Information Processing - ICICIP-2012 The ICICIP 2012 theme was about practical applications of theory and methodologies, to analyze the state-of-the-art developments, open problems and future trends in CS, IT & Computer Applications field. Dr. Chee Peng Lim spoke about computation becoming essential in storing, mining, and analyzing data in modern era. Dr. V Prithviraj spoke about Innovation computing & information processing and applications. There were lectures on "Computational Intelligence: Architectures, Algorithms and Applications" by Dr. Chee Peng Lin, "Engineering Aspects in Current Information Technology Era" by Prof. Dr. P Naghabushan, and "Emotional Intelligence in Artificial Intelligence" by Dr. T Ravi. International Conf. on ICIP 2012 at MEC MOOKAMBIGAI COLLEGE OF ENGINEERING, KALAMAVUR (REGION-VII) Mr. Suresh Thiagarajan, Dr. V Radhakrishnan, 23 March 2012 National Level Technical Symposium TECXPLO 2K12 Dr. M G Venugopalan, Mr. A Subramanian, This event witnessed various activities that boosted technical skills among Dr. M Sekar, and M Chandra Sekaran students. It provided a platform for the brightest minds to showcase their talent and ingenuity. The events such as Paper Presentation and Technical Quiz were organized by Dept. of Computer Science. Quiz session MIND SPORT was conducted efficiently and eighteen teams from various colleges participated in it. (L to R): Mr. Suresh Thiagarajan, Dr. V Radhakrishnan, Dr. M G Venugopalan, Mr. A Subramanian, Dr. M Sekar, and M Chandra Sekaran CSI Communications ions July

44 SPEAKER(S) TOPIC AND GIST P.S.R. RENGASAMY COLLEGE OF ENGINEERING FOR WOMEN, SIVAKASI (REGION-VII) Mr. Sohan Maheswar 20 January 2012: Workshop on Publishing Articles in Wikipedia The speaker explained how to publish articles in Wikipedia to the gathering. He explained how to create an account in Wikipedia, to edit content, and then how to publish new articles on Wikipedia. Mr. Vickranth Navalar Workshop on Publishing Articles on Wikipedia 27 January 2012: Guest lecture on Blue Eyes Technology The guest speaker talked about Blue Eyes Technology and latest trends in software development. Mr. Sundar Rajan During the session 3 February 2012: Guest lecture on Igniting Young Minds to get into Corporate The guest lecturer emphasized on various ways to frame themselves to get into corporate world, how to attend interview, and how to develop their own project. Honoring the Chief Guest SCMS SCHOOL OF ENGINEERING AND TECHNOLOGY, KERALA (REGION-VII) Mr. Jerin Micheal Jose 16 March 2012: Seminar on Android Application The program started with a technical talk on "Android Evolution and Industry focus". In the talk topics like the need for an application functionality, android services, mobile phone operating systems, advantages of Android, naming style of Android versions, and drawbacks of Android were discussed. SSN COLLEGE OF ENGINEERING, TAMILNADU (REGION-VII) Dr. Rajkumar Buyya, Prof. (Dr.) P Anandhan, and Dr. Sriram Rajamani Mr. Jerin Micheal Jose of Neona Embedded Labz taking the lab session April 2012: International Conference on Recent Advances in Computing and Software Systems (RACSS 2012) There were ten invited talks on various areas of emerging trends presented by industrial professionals and academicians at different sessions of the conference. A total of 56 papers were selected for presentation out of 371 papers submissions in four major areas like Software Engineering, Machine Learning, Networks, and Distributed Computing. In addition, two parallel pre-conference tutorials were also arranged in the areas of Cloud Computing and Machine Learning. Dr. Chitra Babu, Dr. Sriram Rajamani, Dr. Rajkumar Buyya, Ms. Kala Vijayakumar, Mr. Veeraghavan Narayanaswamy, Dr. S Ramasamy, Dr. S Salivahanan, and Dr. R S Milton CSI Communications ions July w.cs indi

45 SPEAKER(S) TOPIC AND GIST VALLIAMMAI ENGINEERING COLLEGE, KATTANKULATHUR (REGION-VII) Mr. A S S Victor 14 March 2012: One-day Workshop on Computer Graphics and Animation Mr. Victor started the workshop with history of animation and continued with concepts of production pipeline of 2D, Frame, Storyboard, Script of animation and moved on to Rastering, Animated Clipping, Compositing etc. He spoke about concept of dual camera viewing system of animated object projection and different camera modes and paths. Later he explained 3D animation using MAYA software. (L to R): Mr. R S Vijay Ravikumaran, Mr. A S S Victor, Mr. S Narayanan, and Ms. R Thenmozhi V.M.K.V. ENGINEERING COLLEGE, SALEM (REGION-VII) Mr. K R Jayakumar 9 March 2012: National Conference titled Emphasis in Software Engineering The speaker spoke about how software engineering has evolved over the past few years and the need to emphasis on software engineering for product development and automation. He also highlighted the Technology trends and engineering challenges faced by the industry and advised the students to be a Software Engineer and not to be a Software Coolie. He clearly showcased how good our engineering is and why we need to change from amateur programmer to software professional. He spoke about Four Stages of Learning and improving oneself. National Conference on EMPHASIS IN SOFTWARE ENGINEERING. Dr. R S D Wahidabanu, Principal, Govt. College of Engineering, Salem and Chairperson, CSI Salem Chapter. Following new student branch was opened as detailed below REGION II B. P. Poddar Institute of Management & Technology (BPPIMT), Kolkata - New student branch was inaugurated on 9th June, 2012 in the kind presence of Prof. Dipti Prasad Mukherjee; Dr. Debasish Jana; Prof. (Dr.) Phalguni Mukherjee; Mr. Prashant Verma; Prof. (Dr.) Sutapa Mukherjee, Mr. Soumya Paul and other dignitaries. The inaugural ceremony was followed by a Technical Seminar on Emerging Areas of Computer Application by Mr. Sushanta Sinha. Prof. D P Mukherjee, RVP, Region II floored the audiences with his lecture on Image Processing. The students found their lectures to be very interactive and informational. Later part of the Technical Seminar was concluded by student members presentation on various emerging topics in computer applications like genetic algorithm, swarm intelligence, and other real-time software applications. Please send your event news to csic@csi-india.org. Low resolution photos and news without gist will not be published. Please send only 1 photo per event, not more. Kindly note that news received on or before 20th of a month will only be considered for publishing in the CSIC of the following month. CSI Communications ions July

46 Articles invited on the theme: History of The next issue (August 2012) of the CSI Communications is on the theme: History of IT in India and will attempt to recreate a feel of the past few decades which saw tremendous emergence of IT in India. Articles/Photos are invited from all members towards enriching ing this issue. The following are most welcome: IT in India Brief write up on Computer Centres of s (R&D institutions, Universities, Institutes, Colleges etc) Photographs of early computers and peripherals Interesting anecdotes on early computerisation Interesting computer-related Advertisemnts from media The story of IBM in India (and its exit in late 1970s) History of ECIL Railway and Bank Computerisation Early e-governance efforts Profile of early experts, leaders and researchers in IT with photos and recollections if available CSI Membership = 360 Knowledge Your membership in CSI provides instant access to key career / business building resources - Knowledge, Networking, Opportunities. CSI provides you with 360 coverage for your Technology goals Learn more at Join us and become a member WE INVITE YOU TO JOIN Computer Society of India India's largest technical professional association I am interested in the work of CSI. Please send me information on how to become an individual/institutional* member Name Position held Address City Postal Code Telephone: Mobile: Fax: *[Delete whichever is not applicable] Interested in joining CSI? Please send your details in the above format on the following address. helpdesk@csi-india.org CSI Communications July

47 CSI Calendar 2012 Prof. S V Raghavan Vice President & Chair, Conference Committee, CSI Date Event Details & Organizers Contact Information July 2012 Events 20 July 2012 Hands on workshop on Advanced Excel 2007 CSI Mumbai Chapter July 2012 Hands on workshop on Web Application Security CSI Mumbai Chapter 21 July 2012 Workshop on Developing & Writing Structured Use Cases CSI Mumbai Chapter July 2012 Hands on workshop on Microsoft SQL Server reporting Services (SSRS) CSI Mumbai Chapter July 2012 International Conference on Advances in Cloud Computing (ACC-2012) CSI, Bangalore Chapter and CSI Division I 27 July 2012 Workshop on Group Dynamics and Team Building CSI Mumbai Chapter July 2012 Hands on workshop on e-crime & Computer Forensics CSI Mumbai Chapter 28 July 2012 Workshop on Bidding I.T. Projects; A Successful Approach CSI Mumbai Chapter Mr. Abraham Koshy koshy@csimumbai.org Mr. Abraham Koshy koshy@csimumbai.org Mr. Abraham Koshy koshy@csimumbai.org Mr. Abraham Koshy koshy@csimumbai.org Dr. Anirban Basu, abasu@pqrsoftware.com Dr. C R Chakravarthy, drchakra32@gmail.com Mr. Abraham Koshy koshy@csimumbai.org Mr. Abraham Koshy koshy@csimumbai.org Mr. Abraham Koshy koshy@csimumbai.org August 2012 Events 3-4 August nd AP State Student Convention Dadi Institute of Engineering & Technology, Vizag, A.P 9-10 August August Aug-1 Sep 2012 Regional Student Convention Region 5 G. Pulla Reddy Engineering College, Kurnool, A.P 6th Tamilnadu State Student Convention R.V.S College of Engineering, Dindigul, T.N 3rd International Conference on Transforming Healthcare with IT CSI Division II ( Software), Hyderabad Prof. G Satyanarayana, csi@dielakp.com Dr. K Rajashekhara Rao, krr_it@yahoo.co.in Prof. I K Ishthaq Ahamed, ishthaq@gmail.com T Sabapathy, tspathy@cymphony.com Dr. C G Ravichandran, cg_ravi@yahoo.com Dr. M Sundaresan, bu.sundaresan@gmail.com Dr. T V Gopal, gopal@annauniv.edu September 2012 Events 5-7 September September September 2012 International Conference on Software Engineering (CONSEG 2012) CSI Division II ( Software), Indore Global Science and Technology Forum Business Intelligent Summit and Awards CSI Division II ( Software), Singapore 4th e-governance Knowledge Sharing Summit (KSS2012) Govt. of Chhattisgarh, In association with CSI-SIG-eGOV at Hotel V W Canyon Raipur Dr. T V Gopal, gopal@annauniv.edu Dr. T V Gopal, gopal@annauniv.edu Mr. A M Parial, ceochips@nic.in Maj. Gen. (Retd) Dr. R K Bagga rbagga@iiit.ac.in October 2012 Events 20 Oct Communication Technologies & its impact on Next Generation Computing (CTNGC-2012) I.T.S Management & IT Institute Mohan Nagar, Ghaziabad, U.P November 2012 Events Prof. Umang, umangsingh@its.edu.in Prof. Ashish Seth, ashishseth@its.edu.in Prof. Alka Agrawal, alkaagrawal@its.edu.in 29 Nov- 1 Dec 2012 Third International Conference on Emerging Applications of Information Technology (EAIT 2012) CSI Kolkata Chapter Event at Kolkata URL: D P Mukherjee/Debasish Jana/ Pinakpani Pal/R T Goswami csieait@gmail.com December 2012 Events 1-2 December December December th Annual National Convention of CSI (CSI 2012) CSI Kolkata Chapter Event at Kolkata, URL: Second IEEE International Conference on PDG Computing [PDGC 2012], Technically CSI Special Interest Group on Cyber Forensics at Jaypee University of information Technology, Waknaghat- Solan (HP) International Conference on Management of Data (COMAD-2012) SIGDATA, CSI, Pune Chapter and CSI Division II D P Mukherjee/Debasish Jana/ Pinakpani Pal/R T Goswami, csical@gmail.com Dr. Nitin, pdgc2012@gmail.com Dr. Vipin Tyagi, dr.vipin.tyagi@gmail.com Mr. C G Sahasrabudhe Shekhar_sahasrabudhe@persistent.co.in

48 Registered with Registrar of News Papers for India - RNI 31668/78 If undelivered return to : Regd. No. MH/MR/N/222/MBI/12-14 Samruddhi Venture Park, Unit No.3, Posting Date: 10&11 every month. Posted at Patrika Channel Mumbai-I 4th floor, MIDC, Andheri (E). Mumbai