Evaluation of the Impact of Noise on Iris Recognition Biometric Authentication Systems

Evaluation of the Impact of Noise on Iris Recognition Biometric Authentication Systems Abdulrahman Alqahtani Department of Computer Sciences, Florida Institute of Technology Melbourne, Florida, 32901 Email: aalqahtani2011@my.fit.edu Abstract Iris recognition is a particular type of biometric system that uses pattern-recognition techniques on images of irides to uniquely identify an individual. Here I provide evidence that the iris has a unique pattern which gives us reliable form of identification. In this paper, we will discuss the image quality of irises during iris recognition processing. We used the CASIA-IrisV3-Interval of iris database as input in our system. In this paper, we focused on studying and implementing image quality for each image of CASIA-IrisV3- Interval database by adding noise to that database. We used the Hamming Distant technique (HD) to compare each image on CASIA Database with other images in the same database before and after adding noises. When we get the results Of Hamming distance (HD), we will calculate the TPR and FPR to build ROC Carves for evaluating the image quality of CASIA images. Keywords: iris recognition, Biometric authentication, Iris recognition, image quality, Evaluation of the impact of noise on iris recognition I. INTRODUCTION The use of Iris Biometric is one of the important and successful traits used for the authentication and identification of people for different legal or other critical issues. Initially, this technique was used to recognize an Afghan girl falsely after 18 years with the help of her iris patterns. After that the technique was used widely for authentication and identification for different purposes [1]. There are no two irises which are the same in their mathematical detail even between one s own left and right eye, or between identical twins and triplets [2]. Biometrics systems have come into existence with the revolution of information technology and computer systems. The main purpose of biometrics is to identify individuals and biometrics recognizes the identity of an individual rather than what the person has [3]. Biometrics is not only a system to recognize the identity of an individual, rather it can also be used in our society to reduce fraud, for convenience, and to make our society safer [4]. Biometrics is usually evaluated on the basis of the biometrics recognition performance. The quality and ruggedness of the sensors are two factors that effect the biometrics recognition performance [5]. There are basically two errors in the biometrics system i.e. false rejection rate FRR (rejection of the client) and the false acceptance rate FAR (acceptance of an impostor) [5].There are number of developments that are responsible for the growth of the biometrics industry. e.g. the development of software standards, BioAPI, Biometric Service Providers and Biometric Information Records [6]. There are many challenges for the Iris Recognition System (IRS). Our experiment focuses on studying and implementing iris recognition, comparing each image with others in same database (to get the image quality for each image) by using the Chinese Academy of Sciences - Institute of Automation (CASIA) database for testing our project [7]. In this project, we have a number of sub-systems corresponding to each stage of iris recognition authentication systems. These stages are as follows: Segmentation, Normalization, Feature encoding and Matching. The Results of this IRS are Iris segmentation templates. Iris normalization templates. Polar Mask templates. Polar Noise templates. We used the Hamming Distance (HD) technique to match between the iris images of the CASIA database. The IRS has a great advantage. When we compare the IRS to other visual recognition systems there is huge variability. The Iris pattern between individuals varies to such a degree, that it allows large databases to search with reduced number of false matches [7]. According to recognition of human Iris patterns for biometric identification that was presented by L. Masek, the first step of iris recognition is to isolate practical iris regions in a digital eye image. The imaging qualities of eye imaging will predict the quality of segmentation success. After that, the normalization process produces iris regions which contain the same constant dimension, such that two photographs of the same iris, under different conditions, will have characteristic features at the same spatial locations. In feature encoding and matching, the significant features of the iris must be encoded so that comparisons between

templates can be determined. Most iris recognition systems make use of a band pass decomposition of the iris image to create a biometric template [8]. A. Biometric authentication II. BACKGROUND Biometrics is basically a terminology that has been derived from the two Greek words, bio, which means life, and metrics which means to measure. It is defined as a method to recognize individuals on the basis of their psychological and behavioral methods [3]. This method of biometrics is preferred to the traditional method of passwords and pin numbers because of its accuracy and authenticity [9]. The word biometrics denotes automatic identification of persons grounded on their interactive and organic features. Numerous biological, along with social, biometric individualities have been used. For example, Fingerprint, Palm Print, Face, Iris, Retina, Ear, Voice, Signature, Gait, Hand, Vein, Odor, DNA, etc., are contingent on various kinds of requests. Biometric characters are developed relating satisfactory devices and unique topographies are mined to customize a biometric pattern in registration development. Throughout verification, which is also known as authentication procedure or identification, it can be an identification that is touched as an order of verifications and selections. The organization procedures are the additional biometric dimension which is likened beside the kept pattern(s) producing rejection or acceptance [10]. Historically, humans used faces to recognize each other, but with the increase in populations and as the mode of transport increased, the need for recognition increased, which led to emergence of the field of biometrics. The biometric system can either be a verification system, or an identification system [9]. A verification system is basically to confirm a person s claimed identity, while the identity of a person is established in the identification mode. [9]. Applications of biometrics include mobile phones, secure electronic banking, and computer systems security, secure access to buildings, health, credit cards and social services [3]. Commonly used biometrics includes: Infrared thermogram (hand, hand vein, and facial), Gait which is the peculiar way one walks, Keystroke, ordor, Ear, Hand print, Retina, Iris, Palm print, Voice, Face, Signature, and DNA [3]. There are basically two phases in the biometrics system; i.e. the recognition phase, and the learning phase [5]. An item under consideration is recorded with the help of sensors when the digital data is available. The data is not used directly; rather, some of the data characteristics are extracted from the digital data first to form template [5]. The responsibility of the learning phase is to create a model, e.g. a statistical model. The recognition phase deals with the decision to be taken [5]. The three main functions that biometrics can perform is positive identification of the individual, i.e. Positive Identification (does the system know this person?). It is used for large scale identification, i.e. (whether the person is in the data base or not). Lastly, biometrics perform a function of screening, such that it asks the question, Whether this is a wanted person or not? [4]. Biometrics offer huge amount of security and privacy as compared to other methods of identifying individuals. In some cases the biometrics can be considered as a replacement of the technology [5]. There are basically three main applications of biometrics, i.e commercial, government, and forensic applications. The commercial application includes electronic data security, ATMs, physical access control, Internet access, computer network logins, e-commerce and credit cards, etc. The government applications include correctional facilities, welfare-disbursement, social security, national ID cards, driver s licenses, and passport control, etc. Finally, forensic applications include criminal investigation, paternity determination, terrorist identification, corpse identiïňacation and missing children. The increased threats of terrorism have seen an increased use of biometrics today. A human hand has a combination of different features and they vary significantly from one person to another. Geometric measurement are mostly used as a means of recognitions in commercial systems. Reference pegs are mostly used in geometric measurements for capturing the image of the hand. The most important factor in geometric measurement is a user s acceptability [11]. There is zero risk of the biometrics being lost or forgotten, so the potential threat of intruders is also minimized by the use of the biometrics. Identification is more difficult than verification. Large numbers of comparisons are implementing biometric systems in order to identify individuals. The individual who wishes to remain anonymous can be deprived of their privacy by these biometric systems [5]. Multi-modal biometric systems use the data provided by multiple biometric sources to identify an individual [12]. Information collected from different sources is amalgamated from three levels; i.e., match score level, decision level, and the feature extraction level [12]. B. Iris Recognition In this paper, we have a number of sub-systems that correspond to each stage of the iris recognition. These stages are: - Image acquisition( CASIA-Iris (Chinese academy of sciences-institute of automation) database) Segmentation Normalization Feature encoding

According to the Human Iris Recognition Patterns for Biometric Identification that are presented by L. Masek. Masek s method uses the automatic segmentation system which is established on the Hough transformation method. It can get the location of the circular iris and pupil regions. In the Normalization stage, the extracted iris region (the results of segmentation) is normalized into a rectangular block with constant dimensions to account for imaging inconsistencies. Lastly, to encode the unique pattern of the iris into a bit-wise biometric template, the phase data from 1D Log-Gabor filters must be extracted and quantized into four levels. The Hamming distance is used for the rating of the iris templates. The couple templates are found to match if the testing of statistical independence has been unsuccessful. In our project, the inputs to the system are the eye images, and the outputs are the iris templates [8], [13], [14], [15], [16]. Image Acquisition is taking an image from an iris in the initial stage of an iris-based recognition system [3]. In our project, we do not use this technique because we implemented it with the Chinese academy of sciences-institute of automation (CASIA) database. In fact, an image acquisition captures more than just the iris; e.g. pupil, eyelid, Eyelashes and Sclera (white part of the eye). Also, segmentation isolates the iris from the rest of the eye. There are several techniques and algorithms available Hough transform which was implemented by Wildes et al, which we used in our experiment. Daugman s algorithm (integro-differential operator) [14] [15] It was submitted in 1993. It was the first method efficiently employed on the biometric system. The data-set of this technique is known integro-differential operator [15]. Shrinking and expanding active contour methods [17] These methods are unified when localizing inner and outer iris boundaries. First, the pupil region is assessed based on histogram threshold and morphological operations. Afterward, it uses this data to locate the inner iris boundary. Finally, the inner iris boundary is taken as an initial contour to obtain the outer iris boundary. There are several techniques available for Normalization Daugman s rubber sheet model, we used this technique in our project. Image registration methods. Image registration is processing of more than one image which is used by the same scene taken at different times, from different sides, or by using different sensors [18] [8]. Virtual circles are used to ensure accurate location. [19] The segmented iris region is normalized to eliminate dimensional inconsistencies between iris regions. This is done by implementing a version of Daugman s rubber sheet model [15] [18]. There are several feature extraction and encoding methods and techniques available Log Gabor filters, which we used in our experiment. [13] Zero crossings of 1D wavelet; a wavelet is a function that is used to build a representation [19]. Laplacian of Gaussian, it is used for evaluation of the qualities of the iris images [8] [20] [20]. By using 1D Gabor wavelets with the normalized iris pattern, we are able to prove that Feature encoding is based on the polar coordinate on the 2D normalized iris image. [13]. There are several methods and techniques available for matching Hamming distance which was implemented by Daugman, which we used in our experiment. [8], [14] Weighted Euclidean distance that is employed by Zhu et al. It implemented Weighted Euclidian Distance (WED) (2000) by Zhu et al. It is used to compare the distance between the two templates, particularly if the templates are composed of integer values [21]. Normalized cross-correlation that is employed by Wildes et al, [20] [22]. It is used for matching parts of the images in many applications. Matching methods based on normalized cross-correlation can handle the scale changes between the two images, where there is translation or rotation [22]. C. Image quality There are numerous research projects and papers which discussed the image quality of irises, most of them focused on the noise effects and the different algorithms of iris segmentation. In this paper, we focused on the image quality of iris recognition by comparing each image with other images from the same database, which is CASIA V3 database. As

we mentioned in previous sections, that the image of the eye has to follow all iris recognition stages to obtain the final decision of the iris recognition system (matching or non-matching). Our aim by using an iris recognition system is to know the rate of image quality for each image by comparing it with others. Then, we build our ROC curve according to Hamming distance (HD) techniques. Previous research on iris image quality can be divided into two classes: local and global analysis. Zhu uses the employing discrete wavelet decomposition to measure iris quality by analyzing the coefficients of iris texture [23]. Chan et al have classified the iris quality by evaluating the vitality of concentric iris bands acquired from 2-D wavelets [24]. Ma et al. characterized defocus, motion, and occlusion by analyzing the Fourier spectra of local iris regions [25]. Zhang focused on the sharpness of the area between the pupil and the iris [26]. Daugman and Kang described quality by quantification of the energy of high frequencies over all image regions [27], [13]. Most of the previous research evaluation of iris quality involves some traditional segmentation methods [28]. There are four popular image segmentation methods: Daugman s method Daugman supposed that both pupil and iris circular form applies an intergrodifferential operator. Wildes s method This method executes iris contour appropriate in two steps [29] First step converted the image information into a binary edge map. Second step for particular contour parameter values uses the edge points vote. Mask s method: Using the Kovesi s edge detector, variation is known as Canny edge detector. Then, the next step is applying the circular Hough transform to determine the iris/sclera, and each image correspondent to the iris/pupil. Liam and Chekima method: This technique is based on the fact that the pupil is darker than the iris, but the iris is darker than the sclera. In most cases, the iris images have been taken in less attitudinized imaging conditions, including noise which is localized in some of the iris subparts. Usually, reflections are in the left /right iris and the obstructions are in the upper or lower part of the iris. In the lower and middle-lower signal frequencies, the common feature extraction methods focus on ways to make it more likely that noisy data is included to create the biometric signature. We should divide the iris into different regions in order to detect the regions that are noise-free, then use it to compare with enrolled regions for accurate recognition. Hugo Proenca and A. Alexandre, in their experiment captured Iris images under simulated noncooperative conditions to reduce false rejections. Moreover, they have a free database (UBIRIS), which has distinguishing characteristics of the two free databases (CASIA and UPOL), and the exiting of this data is noise-free. Hugo Proenca and A. Alexandre s experiment has compared the aforementioned four image segmentation methods by using UBIRIS and CASIA [29]. Mayank Vatsa and his group in their experiment proposed that iris indexing algorithms use local and global features to decrease the identification time without compromising accuracy. This algorithm significantly minimizes the computational time without any effect on the accuracy of identification. Mayank Vatsa and his group discussed the challenge of improving performance by comparing the verification and identification algorithms using these iris databases: CASIA Version 3, ICE2005 and UBIRIS [30]. Peihua and Hongwei s experiment discussed the iris recognition problem with errant capture in non-idealistic imaging conditions. In those cases, the iris recognition is challenged by noisy factors; e.g. the off-axis imaging, pose variation, image blurring, illumination change and occlusion, specular highlights, and noise. They provided a robust algorithm based on the Random Sample Consensus (RANSAC) to localized non-circular boundaries. Peihua and Hongwei asserted that these methods can be more accurate than Hough transform methods for localization of iris boundaries by selecting the method based on LucaseCkanade algorithm. According to their experiment, one image could divide into small sub-images and this fixes registration problems for every small sub-image by operating the filtered iris image. Peihua and Hongwei presented their selection method for getting a sub-optimal subset of filters from a family of Gabor filters using UBIRIS.v2 databases. The recognition performance will be greatly improved with a small number of section filters [31]. III. METHODOLOGY. In our experiment, we used the iris recognition system with the CASIA-IrisV3-Interval database then we generated the ROC curve for original database. We added different noises to images of original database for creating different data images with different noises. The ROC curves were generated for each data with its noise. Finally, we compared between ROC curves with and without noises. Overview of our proposed method is shown in Figure 1. A. Data We used CASIA V3-Interval. The Chinese academy of sciences-institute of automation (CASIA) database. There

Figure 1: Overview of our proposed method. Figure 2: The process of adding noise. are a few iris databases which have sample images that are available free of charge to the public. These iris databases share a vast amount of iris images, made in different places. Most of these iris databases, when they are constructed, will strive to have a vast collection of quality iris images. [32]. The Center For Biometric and Security Research in China provided this database to the public in order to improve iris recognition system research around the world [7]. The CASIA-IrisV3 contains three types which are known as CASIA-IrisV3-Interval, CASIA-IrisV3-Lamp and CASIA- IrisV3-Twins. CASIA-IrisV3 includes a total of 22,035 iris images that were taken from more than 700 subjects. All iris images are collected under near infrared illumination, in which all images are 8 bit gray-level JPEG files [7]. In our experiment, we used the CASIA-IrisV3-Interval (Chinese academy of sciences-institute of automation) database. The CASIA-IrisV3- Interval includes a total of 2,639 iris images that were taken from 249 subjects. We used The CASIA-IrisV3-Interval in our experiment because it is the only database for which we have access. The issue with this database is that there are some images which did not work with our system because they are corrupted, or because Mask segmentation rejects them. This mask segmentation problem is one which we faced in our experiment. B. Modification of Image Quality In our implementation, we added different noise to all images of CASIA-IrisV3-Interval to study the quality of images with those noises by following equation. Noise image= Original image + numbers of noise (0.01,0.05,0.09 or 0.1) *randn(size(original Image)) (see Figure 2) Much testing was done to add noise with sub-data. We decide to add those number 0.01, 0.05,0.09 and 0.1 to all images on the database, because we found that they produced clearer results. IV. EVALUATION OF THE IMPACT OF NOISE ON IRIS RECOGNITION In our experiment, we used the Mask method which uses the Kovesi edge detector. Variation is known as Canny edge detector. The next step is applying the circular Hough transform to determine the iris/sclera that is correspondent to iris/pupil [8] This Segmentation approach has the following in common: To detect the contrast between the pupil and the iris for the inner boundary. To detect the contrast between the sclera for the outer boundary, and the approximate distance between the boundaries of the circles around the iris. After segmentation, an iris mask is generated. Automatic segmentation is achieved using the circular Hough transformation method since it is able to localize the circular iris and pupil regions. The iris, pupil regions, and the linear Hough transform, are used for localizing the occluding eyelids. The Threshold was also used for isolating eyelashes and reflections. [8] The next step, after the iris segmentation process, will be iris normalization. In order to allow comparisons, the iris region will transform into fixed dimensions. The purpose of this stage is to generate constant dimensions of the iris regions. The technique implemented is Daugman s Rubber sheet model [8], [14], [15], [18]. Features of the iris are encoded by contortion of the normalized iris region with 1D Log-Gabor filters and phase quantization of the output in order to generate a bitwise biometric template (see Figure 3 and 4). The Hamming distance (HD) was selected as a matching metric that gave a measure of how many bits conflicted between the two templates [8]. Failure of statistical independence between two templates would result in a match; that is, the two templates were deemed to have

Figure 3: Samples template of iris with noise. Figure 5: Comparison of the accuracy result on the CASIA datasets using leave-one-out cross validation. VI. CONCLUSION Figure 4: Samples mask of iris with noise. been created from the same iris if the Hamming distance produced was lower than a constant Hamming distance [8]. When we got the HD results for each Noise data image, we built the ROC curve for each one according to True positive rates (TPR), and False positive rates (FPR), following these equations: TPR = TP P = TP (TP + FN) FPR = FP N = FP (FP + TN) V. RESULTS The higher level of noise is noted to provide less clearer images indicated by the values of the true positive rate. Consequently there appears to be an inverse relationship between the level of generated noise and quality of images. The ROC curve with 0.1 noise produced the lowest TPR value. The ROC curve with 0.01 showed the highest TPR value. The ROC curve with 0.05 has produced an intermediate value with regards to TPR. Images with 0.09 noise produced TRP value slightly better than that of 0.1 noise. The ROC curves demonstrated (Figure 5) that application of different levels of noise influenced the image quality. Images without noise have been shown to produce the clearest results with regards to quality of images in CASIA V3 database. With noise the iris images are giving us significant reliability of images that is still good enough for identification. The iris of a human eye is the most suitable biometric trait to be used in the authentication system, because it provides a unique identification parameter. The purpose of this paper is to prove that the reliability of iris is superior to other biometrics even in the presence of noise that could come with captured iris images. There are some issues which need to be addressed in our work. First, we faced challenges with iris augmentations in which some of the images were not working with our system because of corruption of the images. Also, the automatic segmentation is insufficient, because it could not successfully segment the iris regions for all of the eye images. However, there are many issues in biometric systems such as FMR and FNMR, which can affect the performance of biometric systems, especially if one of them has a high rate. These days, iris recognition systems have been undergone multiple studies and research to increase performance and minimize cost. Iris Recognition is one of the most important and preferable traits used for the identification and authentication of humans for different purposes [33]. The iris has a unique identity for each person. Our system was studying and implementing the various iris recognition schemes available by using Chinese academy of sciencesinstitute of automation (CASIA) database [7] to know the image quality of each image in CASIA V3 database. The major advantage of the iris, is that when compared to other visual recognition techniques, we find that there is a huge degree of variability in the patterns between individuals. This meant that the large databases can be searched without resulting in any false matches [7]. Although the technique was introduced with a high level of security, there was still some chance of compromise. This meant that in spite of high security, there were chances to recover data, which may lead to the unwanted extraction of data. There are other techniques which could be used to make biometrics more secure, called biometric

cryptosystems. The use of Iris Biometric Cryptosystem is one of the most important and successful traits used for the authentication and identification of people for different legal or other purposes [33]. REFERENCES [1] M. Boyd, D. Carmaciu, F. Giannaros, T. Payne, W. Snell, and D. Gillies, Iris recognition, Technical report, Department of Computer Science, Imperial College London, Tech. Rep., 2010. [2] A. Lin, S. Lin, and V. Yen, Eye know you. [3] K. Delac and M. Grgic, A survey of biometric recognition methods, in Electronics in Marine, 2004. Proceedings Elmar 2004. 46th International Symposium. IEEE, 2004, pp. 184 193. [4] A. K. Jain, S. Pankanti, S. Prabhakar, L. Hong, and A. Ross, Biometrics: a grand challenge, in Pattern Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on, vol. 2. IEEE, 2004, pp. 935 942. [5] B. Dorizzi, Biometrics at the frontiers, assessing the impact on society, technical impact of biometrics, European Parliament Committee on Citizens Freedoms and Rights, Justice and Home Affairs (LIBE), Technical Report, 2005. [6] E. S. Dunstone, Emerging biometric developments: Identifying the missing pieces for industry, in Signal Processing and its Applications, Sixth International, Symposium on. 2001, vol. 1. IEEE, 2001, pp. 351 354. [7] L. Li, F. Xu, H. Wang, C. She, and Z. Fan, Chinese academy of sciences, 2004. [8] L. Masek et al., Recognition of human iris patterns for biometric identification, Ph.D. dissertation, MasterâĂŹs thesis, University of Western Australia, 2003. [9] S. Angle, R. Bhagtani, and H. Chheda, Biometrics: A further echelon of security, in UAE International Conference on Biological and Medical Physics, 2005. [10] C. Rathgeb, Iris based biometric cryptosystems, Ph.D. dissertation, Diplomarbeit, Salzburg University, 2008. [11] A. Ross, A. Jain, and S. Pankati, A prototype hand geometry-based verification system, in Proceedings of 2nd Conference on Audio and Video Based Biometric Person Authentication, 1999, pp. 166 171. [12] A. Ross and R. Govindarajan, Feature level fusion in biometric systems, see http://www. wvu. edu/ bknc/2004% 20Abstracts/Feature% 20Level% 20Fusion% 20in% 20Biometric% 20Systems. pdf, 2004. [13] J. Daugman, How iris recognition works, Circuits and Systems for Video Technology, IEEE Transactions on, vol. 14, no. 1, pp. 21 30, 2004. [14] J. G. Daugman, Biometric personal identification system based on iris analysis, Mar. 1 1994, us Patent 5,291,560. [15] P. Verma, M. Dubey, P. Verma, and S. Basu, DaughmanâĂ s algorithm method for iris recognition-a biometric approach, International Journal of Emerging Technology and Advanced Engineering, vol. 2, no. 6, pp. 177 185, 2012. [16] P. D. Kovesi, Matlab and octave functions for computer vision and image processing, Online: http://www. csse. uwa. edu. au/ pk/research/matlabfns/# match, 2000. [17] K. Nguyen, C. Fookes, and S. Sridharan, Fusing shrinking and expanding active contour models for robust iris segementation, in Information Sciences Signal Processing and their Applications (ISSPA), 2010 10th International Conference on. IEEE, 2010, pp. 185 188. [18] T. W. Hsiung and S. S. Mohamed, Performance of iris recognition using low resolution iris image for attendance monitoring, in Computer Applications and Industrial Electronics (ICCAIE), 2011 IEEE International Conference on. IEEE, 2011, pp. 612 617. [19] W. W. Boles and B. Boashash, A human identification technique using images of the iris and wavelet transform, Signal Processing, IEEE Transactions on, vol. 46, no. 4, pp. 1185 1188, 1998. [20] R. P. Wildes, J. C. Asmuth, G. L. Green, S. C. Hsu, R. J. Kolczynski, J. Matey, and S. E. McBride, A system for automated iris recognition, in Applications of Computer Vision, 1994., Proceedings of the Second IEEE Workshop on. IEEE, 1994, pp. 121 128. [21] H. Ali, M. Salami et al., Iris recognition system by using support vector machines, in Computer and Communication Engineering, 2008. ICCCE 2008. International Conference on. IEEE, 2008, pp. 516 521. [22] F. Zhao, Q. Huang, and W. Gao, Image matching by normalized cross-correlation, in Acoustics, Speech and Signal Processing, 2006. ICASSP 2006 Proceedings. 2006 IEEE International Conference on, vol. 2. IEEE, 2006, pp. II II. [23] X.-D. Zhu, Q.-l. Cui, Y.-N. Liu, and X. Ming, A quality evaluation method of iris images sequence based on wavelet coefficients in" region of interest", in Computer and Information Technology, International Conference on. IEEE Computer Society, 2004, pp. 24 27. [24] Y. Chen, S. C. Dass, and A. K. Jain, Localized iris image quality using 2-d wavelets, in Advances in Biometrics. Springer, 2005, pp. 373 381. [25] L. Ma, T. Tan, Y. Wang, and D. Zhang, Personal identification based on iris texture analysis, Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 25, no. 12, pp. 1519 1533, 2003. [26] M. Salganicoff and G. H. Zhang, Method of measuring the focus of close-up images of eyes, Sep. 14 1999, us Patent 5,953,440. [27] B. J. Kang and K. R. Park, A study on iris image restoration, in Audio-and Video-Based Biometric Person Authentication. Springer, 2005, pp. 31 40. [28] N. D. Kalka, J. Zuo, N. A. Schmid, and B. Cukic, Image quality assessment for iris biometric, in Defense and Security Symposium. International Society for Optics and Photonics, 2006, pp. 62 020D 62 020D. [29] H. Proenca and L. A. Alexandre, Toward noncooperative iris recognition: A classification approach using multiple signatures, Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 29, no. 4, pp. 607 612, 2007. [30] M. Vatsa, R. Singh, and A. Noore, Improving iris recognition performance using segmentation, quality enhancement, match score fusion, and indexing, Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on, vol. 38, no. 4, pp. 1021 1035, 2008. [31] P. Li and H. Ma, Iris recognition in non-ideal imaging conditions, Pattern Recognition Letters, vol. 33, no. 8, pp. 1012 1018, 2012. [32] R. Parashar and S. Joshi, Comparative study of iris databases and ubiris database for iris recognition methods for non-cooperative environment, International Journal of Engineering, vol. 1, no. 5, 2012. [33] C. Rathgeb and A. Uhl, The state-of-the-art in iris biometric cryptosystems, State of the art in Biometrics, pp. 179 202, 2011.