Noise Removal in Thump Images Using Advanced Multistage Multidirectional Median Filter

Transcription

1 Volume 116 No , 1-8 ISSN: (printed version); ISSN: (on-line version) url: ijpam.eu Noise Removal in Thump Images Using Advanced Multistage Multidirectional Median Filter 1 Y. Saleena and 2 V.S. Dharun 1 Department of Computer Science and Engineering, Noorul Islam University, Kumaracoil, Tamilnadu, India. saleenashajar@rediffmail.com 2 IAJJ College of Engineering, Marthandam, Tamilnadu, India. dharunvs@yahoo.com Abstract Removing noise in thump images is becoming a new area of research in image processing as thump is becoming a vital and important authentication tool in digital world. Our world is always wants to be smarter. But never becomes smart enough to be smarter. In this smart environment, we mostly depend on chips and biometrics for our authentication, which to some extend meets our requirement. In fact, we never come across the vitality and importance of biometrics especially common biometric factor fingerprints. In this work, we discuss about the replacement of smart cards with biometric factors that symbolise the authentication as well the mobility and safety of smart products. Various types of noise occur while taking the thump impressions and among these impulse noise is becoming an important one. Now a day s finger images are the very important thing to be considered in the processing of digital image as various kinds of noise occur while image capturing, transmission or storage. The main information of the thump image should be stored and preserved in the time of Noise removal. There are different of algorithms are available for obtaining the correct image after removing the noise which is impulsive from the corrupted image. An effective and elegant algorithm is discussed in this paper for the removal of impulsive noise from a highly interrupted thump image. The 1

2 thump authentication process works in four steps. The first step is image acquisition, second deal with feature extraction, then comes noise removal and, then comes pattern matching. Noise removal is carried using advanced multistage multidirectional algorithm. This proposed algorithm works in two steps and they are determining the noised pixel and next is the reconstruction of the corrupted images using the proposed algorithm. The proposed algorithm gives very good results than the Standard Median Filter (SMF), Center Weighed Standard Median Filter (CWSMF), Adaptive Standard Median Filter (ASMF), Adaptive Center Weighed Standard Median Filter (ACWSMF) and Directional weighed Standard Median Filter (DWSMF).Peak Signal-to-Noise Ratio (PSNR) is better for the proposed algorithm as compared to methods discussed. Key Words:Standard Median Filter (CNF), Center Weighed Standard Median Filter (CWSMF), Adaptive Standard Median Filter(ASMF), Adaptive Center Weighed Standard Median Filter (ACWSMF) and Directional weighed Standard Median Filter (DWSMF). 2

3 1. Introduction Fingerprint recognition and identification has a main role in biometric authentication because of its simplicity in getting the images. Unwanted information that corrupts an image known as noise will be there in the acquisition process of finger prints. While acquiring finger prints it is a usual scenario of the introduction of impulse noise. So, for efficient authentication of finger prints the noise occurred while taking the image must be removed. Already there are lot of algorithms for noise extraction and restoration of the image. But all are having their own advantages and d is advantages. The main type of noise occurring here are the impulsive noise and they are of two types. They are random valued noise and salt and pepper noise. Median filter available are having a very important and prominent place compared to other methods and technologies available. 2. Background Work As the needs arises various nonlinear filers have been developed for the process of removing impulsive noise. For this a number of nonlinear filters has been developed and proposed for image restoration process which is interrupted by different types of noise. Standard Median Filter (SMF), Center Weighed Standard Median Filter (CWSMF), Adaptive Standard Median Filter (ASMF), Adaptive Center Weighed Standard Median Filter (ACWSMF) and Directional weighed Standard Median Filter (DWSMF) are some of them. Standard Median Filter (SMF) is the easiest and simplest form of median filter for the restoration of image after filtering without making any damage to the edge details. But it is very clear that it will not work in high density noise levels because in these cases the edge details are not preserved. This median filter is good only for lower density levels. In Center Weighed Standard Median Filter (CWSMF) in the filtering window weights are given only to some of the pixels which will control and coordinate the filtering behavior. The main problem that arises is that is that they are not considering the error and corruption in the pixel elements. Yet another problem with this algorithm is that they may not work with high noise densities. So, to overcome the defects the yet another filter known as, Adaptive Standard Median Filter (ASMF) is developed. But it lacks to preserve the main and needed details of the edges. There is another filter known as Adaptive Center Weighed Standard Median Filter (ACWSMF) and it rectify some of the problems. In Adaptive center weighted Standard median filter(acsmf), the main disadvantage is that it need some intensity values known as the threshold values [1], [4]. In the case of Directional weighed Standard Median Filter (DWSMF) the problem is that as the window size increases it becomes effective at low density levels. All the algorithms discussed here fail at high noise level densities. The suggested and proposed work is the enhancement done on the removal of salt and pepper noise at high level noise densities [10] from digital images. The remaining portion of the 3

4 paper is structured as follows. The discussed algorithm is mentioned in Portion III. Simulation results with different images are presented in Portion. Results and comparison is discussed in Portion V. Conclusions are given in Portion VI. 3. Proposed Work The proposed work can be diagrammatically represented as the following diagram. The first step in thump authentication is capturing the thump image, then in the next step the features in the image is extracted. Then comes the main step of noise removal. Impulse noise causes error in data transmission. The interrupted pixels are given highest value by positive impulse or zero is given for negative impulse. This gives a salt and pepper appearance. The unaffected pixels are kept unchanged. Usually the noise is identified by the pixel percentage of corrupted image. If 50% of the image is corrupted then half of the image pixels gets corrupted by positive impulses and other half is corrupted by negative impulse. Most median based filters use median value or weighted median value from the whole window as estimated value. Edge details cannot be recovered due to the replacement of noisy pixels with the vicinity of median value, without considering the presence of edge features. Figure 1: Steps in Noise Removal in Thump Images By sorting all the necessary pixel intensity values of the neighbour pixel the median value can be calculated by rearranging pixel intensity values and then taking the median intensity value as the middle pixel value. (The average of two middle pixel value is determined if the neighbourhood pixel considered has an even number). Here the algorithm works by calculating the median values of intensity along different directions. So, depending the number of directions taken we get a number of values of intensities. Among these median values taken and consider the direction of which we obtained the least standard deviation. The median intensity value for the direction is taken as the standard intensity and this value is given to corrupted pixel. 4. Simulation In the experiment, the original test images are corrupted with the impulsive noise i.e. salt-and-pepper noise. In order to calculate the restoration performance, the Peak signal-to-noise ratio (PSNR) is used and it can be 4

5 represented as, where r and o represent the pixel values of the retrieved image at(i,j) and the source image at(i,j) respectively, and the size of image M x N. 5. Results and Comparison The result analysis of the suggested algorithm is considered with various gray scale thump images for varying noise densities. Here noise densities vary from ten percentage to 90 percentage for thump images. Denoising performances are quantitatively measured using Peak Signal to Noise Ratio(PSNR). The PSNR values of the suggested algorithm is analyzed with the performance and efficiency of other algorithms by varying the noise density from ten percentage to ninety percentage for an image were depicted as in Table 1. From the Table 1, it is evaluated that the efficiency of suggested algorithm is found realistic in low and high noise densities. Table 1: Comparison of PSNR Values of Various Algorithms for the Image of Different Noise Densities Noise in % SMF CWSMF ASMF ACWSMF DWSMF AMMD 10% % % % % % % % % The Leftover Impression which gives Imperfect Distorted Image Image Set for Filtration Adding Salt & Pepper 5

6 Various Stages of Filtration to Remove Noise Edge Removal from Noise Image Resulted Image Various Stage of Pattern Matching 6. Conclusion Authenticated Image The qualitative analysis of the thump images by applying the discussed work is done and it is anonymously clear from the result given in the table that noise is removed effectively and efficiently as compared to other filtering techniques such as Standard Median Filter (SMF), Center Weighed Standard Median Filter (CWSMF), Adaptive Standard Median Filter (ASMF), Adaptive Center Weighed Standard Median Filter (ACWSMF) and Directional weighed Standard Median Filter (DWSMF). It is very clear from the work that noise while capturing the thump images are effectively removed using the proposed method. So, it can be considered in the field of thump authentication process. Advanced Multistage Multidirectional Median Filtering Algorithm does not require any other threshold parameter and also it works good with lengthy processing of images in thump images. Also, it need only the minimum number of iterations. It gives a quite stable performance when the experiments are done. This method is very efficient to eliminate high density noises in the thump recognition system. 6

7 References [1] Ko S.J., Lee Y.H., Center weighted median filters and their applications to image enhancement, IEEE Transactions on Circuits Systems 38(9) (1991), [2] Hwang H., Haddad R.A., Adaptive median filters: new algorithms & results, IEEE transactions on image processing 4 (1995), [3] Astola J., Kuosmanen P., Fundamentals of nonlinear digital filtering, CRC press (1997). [4] Chen T., Wu H.R., Adaptive Impulse Detection Using Center Weighted Median Filters, IEEE Signal Processing Letters 8(1) (2001). [5] Zhang S., Karim, M.A. A new impulse detector for switching median filters, IEEE Signal Processing Letters 9(11) (2002), [6] Gonzalez R.C., Woods R.E., Digital Image Processing, Second Edition (2007). [7] Srinivasan K.S., Ebenezer D., A New Fast and Efficient Decision- Based Algorithm for Removal of High-Density Impulse Noises, IEEE Signal Processing Letters 14 (2007), [8] Nair M.S., Revathy K., Tatavarti R., Removal of Salt-and Pepper Noise in Images: A New Decision-Based Algorithm, Proceedings of the International Multi-Conference of Engineers and Computer Scientists (2008). [9] Esakkirajan S., Veerakumar T., Subramanyam A.N., Prem Chand C.H., Removal of High Density Salt and Pepper Noise Through Modified Decision Based Unsymmetric Trimmed Median Filter, IEEE Signal Processing Letters 18(5) (2011). [10] Benazir T.M., lmran B.M., Removal of High and Low Density Impulse Noise Using Modified Median Filter, International Conference on Recent Trend in Engineering & Technology (2012),

8 8