Touchless Region based Palmprint Verification System

International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 Touchless Region based Palmprint Verification System Satya Bhushan Verma, Saravanan Chandran resolution imaging, low cost hardware, stable structural feature, fast feature extraction, easy availability, high verification accuracy, and high user acceptability. Most of the palmprint recognition approaches are using natural light for image acquisition for biometric applications. Palm is an inner surface of hand and the palm contains three types of lines those are called flexion creases, secondary creases, and ridges. The flexion creases are known as principal lines and the secondary crease are called wrinkles. These lines are used for forming biometric features in palm print verification. Principle lines are used in few palmprint recognition models and ridges with minutiae are used in few palmprint recognition systems. Researchers have been working on high resolution images as well as low resolution images for biometrics applications. The high resolution images mostly used for criminal detection and forensic purpose and low resolution images are mainly used in automatic attendance, gate entry, and public authentication applications. Usually high resolution image are considered as minimum of 400dpi and the low resolution image considered as maximum of 150dpi. It is noticed from the literature in general the palmprint-based identification techniques have been developed using any one of the techniques, principal line based techniques or subspace based techniques or statistical techniques. The principal line based approaches are developed by using edge detection algorithms. In the subspace based approaches, the palmprint image is measured as data with high-dimension, which can be plotted into lower dimensional space. Using the correlation better matching is obtained [7]. In the statistical based approaches series of different filters are applied on palmprint images then the image is determined with a binary scheme. Statistical based approaches proved high verification accuracy, which is most effective recognition method compared to other approaches. Still, various research works are taking place which are focusing on development of efficient palmprint recognition techniques. Abstract— The Biometrics systems have been used for identification of a person with their physical structure or a behavioural characteristic. Palmprint is popular nowadays for identifying a person because of its ridges is big in size compared to fingerprint. A palmprint consists of three types of lines those are called principal lines, secondary lines, and wrinkles. These lines contain rich and robust information of a person which are used as a biometric. Palmprint verification systems are available in touch and touchless systems. The touchless system is preferable than touch based system for easy maintenance. In this research article, an effective touchless palmprint verification system is proposed for identification of a person. The proposed new model is developed by using Gabor filter and Local Binary Pattern (LBP) with their variant that is LBP-U, LBP-RI, and LBP-RIU. The experiment is carried out by using IITD palmprint database and CASIA palmprint database. The Gabor filter based on Rotation Invariant Uniform Local Binary Pattern (GFLBP-RIU) feature extraction technique has produced best total success rate (TSR) is equal to 99.25% and 99.00% in IITD palmprint database and CASIA database respectively. The proposed method takes 0.88 second for palmprint matching. Index Terms— Palmprint, Gabor Filter, Local Binary Pattern, LBP, Biometric. I. INTRODUCTION T HE Biometric techniques have been commonly used for verification of the identity of an individual with their physical structure or behavioral characteristics. The physiological characteristics such as iris, fingerprint, palmprint, face, ear etc. and the behavioral characteristics like as signature, keystroke, gait etc. are unique and used in biometric applications. The biometrics techniques are classified into touch and touchless based techniques. The touchless based techniques are highly preferred for it easy maintenance. The palmprint based biometric techniques is easy implement using touchless approach and its many advantages over other approaches like it works in low ————————————————  Satya Bhushan Verma is currently pursuing PhD in Computer Science from National Institute of Technology, Durgapur, West Bengal, India. E-mail: [email protected]  Saravanan Chandran has completed Ph.D. from Department of Computer Applications, National Institute of Technology, Tiruchirappalli, in 2009. He is currently working as Assistant Professor at the National Institute of Technology, Durgapur, India E-mail: [email protected] II. RELATED WORKS A.H. M. Al-Melali et al. proposed a fast personal palmprint authentication based on 3-D multi wavelet transformation [1]. In that paper they used 3-D discrete multi wavelet transformation as the feature extractor and a probabilistic artificial neural network as a classifier. They tested and 181 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 evaluate their proposed method upon 240 palmprint images. They achieved 100% recognition rate. Wai Kin Kong, David Zhang and Wenxin Li [24] used 2-D Gabor filter for palmprint feature extraction. The palmprint image was captured using a digital scanner. The coordinate system was set based on the boundaries of fingers for extracting the ROI. Further they applied 2-D Gabor filter for textured feature extraction and distance matching. After the experiment they found filter 11th filter is the best of 12 filters in term of accuracy. Dewi Yanti Liliana et al. proposed a biometric palm recognition system combining palmprint and hand geometry [5]. First, they applied preprocessing then they extracted two features one for palmprint and another one for hand geometry. After extracting features from hand geometry and palmprint, they matched with the database by test feature. They tested the model with 100 samples by using three methods, using a) palmprint, b) hand geometry and c) combination of palmprint and the hand geometry. They achieved higher accuracy rate 89% by the combination of palmprint and the hand geometry compared to others. M. I. Ahmad et al. proposed a palmprint recognition scheme by using local feature and global features of palm [11]. They applied DCT co-efficient and LDA co-efficient in local region to obtain a matching score. They applied LDA co-efficient in Global region to obtain another matching score. A fused score is developed using matching score of global region and matching score of local region for making decision. They achieved Genuine Acceptance Rate (GAR) 98% and False Acceptance Rate (FAR) 0.1%. Kamlesh Tiwari et al. proposed a palmprint based recognition system which is based on palmprint based recognition system using local structure tensor and force field transformation [9]. First, they found region of interest in palm image then applied preprocessing. Further, they extracted features in segmented palm image. They used IITD, CASIA, and PolyU palmprint database for performance evaluation. They achieved 99.89% accuracy in CASIA database, 99.97% PolyU database and in IITK database, they found 100%. The kernel value 5 has achieved high accuracy for all three databases. Nagesh Kumar et al. [15] proposed a multimodal biometric system by the combination of face image and palmprint. They extracted feature by using canonical form based PCA method and used Euclidean distance as a distance parameter. The final result was made by the fusion of score level architecture of palm and face and resulted in getting 97% accuracy for the Face, 96% accuracy for the palmprint, and 97% accuracy on the fusion of face and palmprint. Saravanan Chandran et al. presented a Touchless palmprint verification system using shock filter, SIFT, I-RANSAC and LPD (Local Palmprint Descriptors) [18]. Firstly they applied shock filter in segmented palmprint image and showed that the preprocessed image gives more SIFT points than without shock filter. Then they extracted SIFT features and moreover refined the SIFT feature by using I-RANSAC and LPD. This resulted in the preprocessed image giving better matching score than without preprocessing. They used IIT Delhi and CASIA palmprint database. Quan Wang has published a paper and described the Kernel Principle Component Analysis and its application in face recognition [17]. Kernel PCA allows nonlinear dimensionality reduction. In this proposed paper they firstly construct the Kernel Matrix for the data set (test image) and then compute the Gram matrix. Then they calculated the Kernel Principal Component by using vectors. Error rates of 8.82% and 23.08% in the training data and testing data simultaneously in Principal Component Analysis, and 6.86%, 11.54% error rate in the training data and testing data simultaneously Kernel Principal Component Analysis. Ruifang Wang et al. [16] proposed high-resolution palmprint recognition by using spectral minutiae. They divided the palm into three regions called interdigital region, thenar, and hypothenar. The interdigital region contains significant singular points and heart line. The thenar region contains minor creases and wrinkles and hypothenar region contains regular ridges. They applied spectral minutiae representation into each region, made a regional fusion of them, and to get a score. They achieved 2.4% EER for sum rule fusion, and 1.77% for logistic regression based fusion. Slobodan Ribari et al. [19] proposed an approach based on Gabor filter for color palmprint images. First, they decomposed color image into R, G, and B components then applied Gabor filtering and thresholding in each component. They used generalized Hamming distance as a classifier. They achieved recognition accuracy approximately 98.71%. Murat Ekinci et al. proposed a palmprint recognition system using kernel PCA of Gabor feature [14]. In this article, they integrated the Gabor wavelet and kernel PCA methods. The Gabor wavelet first derives necessary palm features by the spatial frequency, spatial locality, and the orientation selectivity. Further, they used kernel PCA method to project the palmprint from the high dimensional palmprint space to a considerably lower-dimensional feature space. Moreover, they calculated Euclidean distance and they used nearest neighbour classifier for features matching and classification. They achieved correct recognition rate 97.22%. Meiru Mua et al. proposed a palmprint identification by using complex directional wavelet and LBP (local binary pattern) [10]. First they extracted the shift and grey scale invariant local features by combining the shiftable (CDFB) Complex Directional Filter Bank Transform and (LBP) Local Binary Pattern. They used Fisher Linear Discriminant (FLD) analysis classifier for palmprint identification. They used Hong Kong PolyU palmprint database for the experiment. They achieved highest accuracy of 99.32 % in 0.022 seconds. A. Wincy and G. C. Chandran, proposed a palmprint scheme by using PCF and SURF feature [3]. Firstly they applied preprocessing in segmented palmprint images. Then they extracted features by using Phase-Correlation Function (PCF) and matched with the database. Further they applied Speeded Up Robust Features (SURF) extraction and matching. They used Hong Kong PolyU palmprint database. They achieved EER of 6.488%. 182 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 ωyο = Centre frequency of y direction in which the filter produces the greatest response. σx = Standard deviation of Gaussian function of x directions. σy = Standard deviation of Gaussian function of y directions. For the real part of Gabor filter above figure 1 is defined in equation 2 [19]. Azadeh Ghandehari and Reza Safabakhsh proposed a comparison between principal component analysis and adaptive principal component extraction for the palmprint recognition [2]. Firstly they applied improved PCA for palmprint recognition and they also used Adaptive principal component extraction. They have also shown the mathematical model of PCA and APEX. They used Euclidean distance and Hamming distance as classifiers. They achieved highest recognition rate of 94.57% in PCA and 98.33% in APEX, by using the Euclidean distance. They achieved highest recognition rate of 95.91% in PCA and 98.67% in APEX, by using the Hamming distance. These various research works encouraged to develop a new better touchless palmprint based biometrics technique which is mentioned in detail in the following section 4. Ψ(x, y, ω, θ) = ω √2πk e ω2 (4x2 +y2 ) 8x2 − x2 (eiωx − e− 2 ) (2) where, xι=(x-xo)cosθ +(y-yo)sinθ, yι=-(x-xo)sin θ+(y-yo)cosθ, (xo, yo) = The center of function. ω = The radial frequency in the radians per unit length. Θ = The orientation of Gabor function in radians. k = √(2ln 2) ((2δ +1)/ (2δ -1)) III. GABOR FILTER AND LOCAL BINARY PATTERN Local Binary Pattern is a visual descriptor which is used in pattern recognition and computer vision as a texture descriptor. Initially, Local Binary Pattern introduced by the Timo Ojala et al. [21, 22, and 23] has been used as a main texture analyzer for analysis of images mainly for its representation of discriminative information. These are some other zones where LBP works effectively such as face recognition, demographic classification, object identification etc. The Gabor filter is a powerful tool in the fields of computer vision and pattern recognition. The Gabor filter is a renowned isotropic filter. It gives many advantages like variation of rotation, translation and illumination, which is raised by capturing the device and palm structure. Gabor filter gives higher flexibility in the definition of function shape, because of more general set of degree of freedom. The flexible size for a circular neighbourhood was proposed in [22] for controlling the insufficiency of the original LBP operator of 3 × 3 neighbourhood size that not able to trace the dominant texture features in to the large scale structures. The LBP label for a centre which has pixel coordinate (x, y) of an image is given by [22]. P (3) LBPP,R (x, y) = ∑P−1 P=0 s(g p − g c ) 2 where, g c = Grey value of pixel of interest (Central Pixel) g p =Neighbourhood of central pixel. 0, z < 0 S(z) = { Function of symbolizes thresholding. 1, z ≥ 0 P= Number of specimen points on circular neighbourhood. R= Spatial resolution of the neighbourhood. Bilinear exclamation is applied to the pixel values if the specimen points are not part of the integer coordinates. In Local Binary Pattern, if there is at most 2 bit wise conversion that is 0 to 1 or 1 to 0, is reported in the circular binary pattern then the LBP is called as LBP uniform pattern. The histogram of LBP-U contains separate bin for the uniform patterns and for all other non-uniform patterns assign only single bin. For the assumed pattern of P bits, the output bins produced by P(P−1)+3. The LBP patterns of natural images are mostly uniform and hence reduce the non-uniform patterns from the images. Uniform patterns in any texture images are account for approx. 90% of the entire pattern with the (8, 1) neighbourhood and close to 70% for the (16, 2) neighbourhood [13]. Fig. 1. (A) Real parts of Gabor filters, (B) Magnitudes of Gabor Filters. Gabor Filter Bank: Gabor filter contains Gaussian function which is modified by the complex sinusoidal of frequency domain as shown in equation 1. G(x, y) = e (x−x0 )2 (y−y0 )2 − j(ωx0 x+ωy0 y) 2σ2 x 2σ2 y − e (1) Where: x, y = Coordinate of pixel position in to image. ωxo = Centre frequency of x direction in which the filter produces the greatest response. The LBP codes may be changed when the image is rotated. The LBP-RI has been proposed in [6, 12 and 23] for this issue. The 183 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 LBP-RI generated by circularly rotation of basic LBP code and considers the pattern with the minimum value. LBP − RIP,R = mini [ROR(LBPPR i)] Fig. 3. Sample palm images from CASIA palmprint database, (a) Hand image (b) segmented ROI respectively. (4) In this research article, in the verification phase the distance between two histograms are calculated by Chi-square distance, Manhattan Distance, and Bhattacharyya distance. where i= 0, 1, 2, -,-,-, P − 1. ROR(x, i) = Circular bit-by-bit right shift operation is performed on x (a P-bit number) for the i times. The LBP − RIP,R descriptor produces 36-bin histograms for each image due to the 36 diverse and 8 bit rotation invariant codes. Rotation Invariant Local Binary Pattern has some disadvantages due to crude quantization of angular space at the 45◦ LBP − RIUP,R was proposed by T. Mäenpää in the year 2003[20]. The Chi-square distance is defined as D(H1 H2 ) = ∑I (H1 (I)−H2 (I))2 (5) H1 (I) The Manhattan distance between two histogram H1 and H2 is defined as j j D(H1 H2 ) = ∑N j=1|H1 − H2 | IV. PROPOSED METHODOLOGY (6) The Bhattacharyya distance between two histogram H1 and H2 is defined as IIT Delhi palmprint image database contains of left and right hand color images of 230 persons in the age group between 14–56 years [8]. Six palmprint samples of both hands have been taken from each person. All the hand images of this database had been taken by contact less CMOS camera which is saved in JPG format. The segmented image of both hand images are also provided in that database which is stored in BMP format at dimension 150X150 at gray scale. These segmented palmprint images are used for experiment. Figure 2(A) shows some sample images from IIT Delhi Touchless palmprint database and figure 2(B) shows segmented ROI respectively. D(H1 H2 ) = √1 − 1 ∑ H (I) ̅ 1H ̅ 2 N2 I 1 √H ∙ H2 (I) (7) where, N=Total number of Histogram bins. FAR (False Acceptance Rate), FRR (False Rejection Rate), TSR (Total Success Rate) and EER (Equal Error Rate) has been used for evaluating the proposed method for verification. In any biometric scheme, the FAR determines the rate of invalid persons who are incorrectly accepted, while FRR determine the total rejection rate for the right persons. The TSR (Total Success Rate) determine the correctness of any biometric system while determine total error in any biometric system. FAR (False Acceptance Rate), FRR (False Rejection Rate), and TSR (Total Success Rate) are used as evaluation standard, they are defined in the following equations 8, 9, and 10. FRR = Fig. 2. Sample palm images from IIT Delhi Touchless palmprint database (a) Hand image (b) segmented ROI FAR = CASIA Palmprint database contains 5,502 palm images which is collected from 312 persons [4]. They collect both left and right palm images from each person. All palm images are in 8 bit gray-level and stored in JPEG format. They do not use any pegs to restrict postures and positions of palm images. Figure 3 (A) are some sample images from the CASIA palmprint database and figure 3(B) shows the segmented ROI respectively. TSR = (1 − NFR NEA NFA NIA X100 % X100 % FAR+FRR TNA ) X100 % (8) (9) (10) Where, NFR= Number of false rejection, NEA=Number of Enrollee Attempts, NFA=Number of False Acceptance, NIA=Number of Impostor Attempts, TNA=Total Number of Attempts. The segmented palmprint region is divided into 9 subregions namely PR1 to PR9 as shown in figure 4. By analysing all the nine sub-regions of the palmprints it is noticed that the three principal lines are found in eight regions PR1, PR2, PR4, PR5, PR6, PR7, PR8, and PR9 but not found in region PR3. Regions PR4, PR5, PR7, and PR8 highly significant in matching process and these regions are used four times in matching process. Further notice, the regions PR1, PR2, PR6 184 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 PR7, and PR8 regions, and image P4 consists of regions PR4, PR5, PR7, and PR8. Gabor filter is applied on P1, P2, P3, and P4 images as shown in figure 5 (C). LBP histogram is applied on G1, G2, G3, and G4 images as shown in figure 5 (D). The figure 2(E) is the cumulative histogram LBP of LBP1+LBP2+LBP3+LBP4. The LBP is the Palmprint feature extracted from the various Palmprint regions. and PR9 are significant in matching process. So these regions are used two times for matching. The remaining insignificant region PR3 is used only once in the matching process. The following figure 5 shows the flow of the Palmprint feature extraction process of the newly proposed technique. P1 is the source segmented palmprint image. Image P2 consists of PR4-PR9 regions, image P3 consists of PR1,PR2, PR4, PR5, Fig. 4. (a) Original segmented image, (b) Divided in 9 sub-region PR1-PR9. Fig. 5. Process of Palmprint feature extraction the Rotation Invariant Uniform Local Binary Pattern when Bhattacharyya distance parameter is used. The highest TSR=99.25% and EER=0.75% is obtained at the Rotation Invariant Uniform Local Binary Pattern by using Bhattacharyya distance parameter. V. RESULT AND ANALYSIS The following table 1 shows the experiment results of the proposed model in FAR, FRR, TSR, and EER by using IIT Delhi Touchless palmprint database at different distance parameters and at all three variant of LBP. The proposed model has achieved highest FAR=1.5% and FRR=1.5% at 185 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 Table 1. Comparison of FAR, FRR, and TSR of the proposed model using IIT Delhi palmprint database Distance parameter Chi-square distance Manhattan distance Bhattacharyya distance LBP Variants FAR FRR TSR LBP-U 5.0% 2.0% 98.25 % LBP-RI 4.0% 3.5% 98.20 % LBP-RIU LBP-U LBP-RI 2.0% 4.5% 5.0% 2.0% 3.5% 2.0% 99.00 % 98.00 % 98.25 % LBP-RIU LBP-U LBP-RI LBP-RIU 2.5% 2.0% 2.5% 1.5% 3.5% 5.5% 5.0% 1.5% 98.50 % 98.20 % 98.20 % 99.25 % Table 2. Comparison of FAR, FRR, and TSR of the proposed model using CASIA palmprint database Distance parameter Chi-square distance Manhattan distance Bhattacharyya distance LBP Variants FAR LBP-U LBP-RI LBP-RIU LBP-U LBP-RI LBP-RIU LBP-U LBP-RI LBP-RIU 6.0% 4.0% 1.5% 4.5% 5.5% 2.0% 4.5% 2.5% 2.0% FRR 3.0% 4.5% 2.5% 3.0% 4.5% 2.5% 6.0% 3.5% 3.0% TSR 97.75% 97.87% 99.00% 98.12% 97.50% 98.87% 97.37% 98.50% 98.75% biometric technique for palmprint verification system using Gabor filter, and LBP (Local Binary pattern) is proposed, conducted an experiment, and discussed the results in detail in this research article. In this article LBP and its variants that are LBP-U, LBP-RI, and LBP-RIU are used. The proposed model is tested in the experiment by using IIT Delhi touchless database and CASIA palmprint database. The proposed model achieved the highest FAR=1.5%, FRR=1.5%, TSR=99.25% and EER=0.75% at the Rotation Invariant Uniform Local Binary Pattern using Bhattacharyya distance parameter in the IITD palmprint database and the highest FAR=1.5%, FRR=2.5%, TSR=99.00% and EER=1.00% at the Rotation Invariant Uniform Local Binary Pattern at Chi-square distance parameter in the CASIA palmprint database. The proposed method takes 0.88 seconds for palmprint verification which is quite fast techniques. Thus, the experiment of the proposed model assures that the proposed model is more suitable for real time palmprint based biometric applications. The above table 2 shows the experiment results of the proposed model in FAR, FRR, TSR, and EER of CASIA palmprint database at different distance parameters and at all three variant of LBP. The proposed model has achieved highest FAR=1.5% and FRR=2.5% at the Rotation Invariant Uniform Local Binary Pattern when Bhattacharyya distance parameter is used. The highest TSR=99.00% and EER=1.00% is obtained at the Rotation Invariant Uniform Local Binary Pattern by using Chisquare distance parameter. The experiment of the proposed model is carried out by using MATLAB 2014a on desktop computer with Intel® core i5-4690 3.5 GHz processor and 4 GB RAM. The proposed method takes 0.88 second for palmprint verification which is quite fast compared to other techniques. The experiment results confirm that the proposed model is more suitable for real time palmprint verification than other models. VI. CONCLUSION The touchless biometric applications are highly preferrable compared to the touch based biometric applications in terms of hygiene and cleaning the surface of sensors, time, speed, cost, etc. Thus, a new touchless 186 https://sites.google.com/site/ijcsis/ ISSN 1947-5500 International Journal of Computer Science and Information Security (IJCSIS), Vol. 15, No. 4, April 2017 patterns, IEEE transactions on pattern analysis and machine intelligence , 24, PP 971–987, 2002. [23] Topi Mäenpää, The Local Binary Pattern Approach to Texture Analysis: Extensions and Applications, University of Oulu, Infotech Oulu, 2003. [24] W.K. Kong, D. Zhang, and W. Li, Palmprint feature extraction using 2-D Gabor filters, Pattern Recognition, 36, 2339-2347, 2003. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] A. H. M. Al-Helali, W. A. Mahmmoud, H. A. Ali, A Fast personal palmprint authentication Based on 3d-multi Wavelet Transformation, International Journal of Scientific Knowledge, vol.2, No.8, 2012. Azadeh Ghandehari and Reza Safabakhsh, A Comparison of Principal Component Analysis and Adaptive Principal Component Extraction for Palmprint Pattern Recognition. Hand-Based Biometrics (ICHB) IEEE, pp 1–6, 2011. Anne Wincy and George Chellin Chandran, Palmprint Recognition using PCF and SURF, International J. of Advanced Research in Com. Science and Software Engineering, V 3, I 10, p996-1001, 2013. ‘CASIA Palmprint Database’, http://biometrics.idealtest.org/dbDetailForUser.do?id=5 Dewi Yanti Liliana and Eries Tri Utaminingsih, The combination of palm print and hand geometry for biometrics palm recognition, International J. of Video & Image Processing and Network Security, V. 12, 2012. G. Zhao, T. Ahonen, J. Matas, M. Pietikainen, Rotation-invariant image and video description with local binary pattern features, IEEE transactions on image processing, 21 1465–1477, 2012. H. Danfeng, Wanquan Liu, Xin Wu, Zhenkuan Pan, and Jian Su"Robust palmprint recognition based on the fast variation Vese– Osher model", Neurocomputing, 2016. IIT Delhi Touchless Palmprint Database, http://www4.comp.polyu.edu.hk/~csajaykr/IITD/Database_Palm.ht m. Kamlesh Tiwari, DevendraK.Arya, G.S.Badrinath, and PhalguniGupta, Designing palmprint based recognition system using local structure tensor and force field transformation for human identification, Neurocomputing, 116, pp222–230, 2013 Meiru Mua, Qiuqi Ruan, and Song Guo, Shift and gray scale invariant features for palmprint identification using complex directional wavelet and local binary pattern, Neurocomputing 74, pp-3351–3360, 2011. M. I. Ahmad, M. Z. Ilyas, R. Ngadiran, Mohd Nazrin, and S. N. Yaakob, Palmprint Recognition Using Local and Global Features, IWSSIP, 2014. M. Pietikäinen, T. Ojala, Z. Xu, Rotation-invariant texture classification using feature distributions, Pattern Recognition . vol. 33, pp 43–52, 2000. M. Pietikäinen, A. Hadid, G. Zhao, T. Ahonen, Computer Vision Using Local Binary Patterns. Computational Imaging and Vision, Computational Imaging and Vision, 2011. Murat Ekinci and Murat Ekinci, Palmprint Recognition Using Kernel PCA of Gabor Features, 23rd International Symposium on Computer and Information Sciences IEEE, 2008. Nageshkumar M., Mahesh P., and Swamy M., An Efficient Secure Multimodal Biometric Fusion using Palmprint and Face Image, The International Journal of Computer Science Issue, vol. 1, pp. 49-53, 2009. Ruifang Wang, Daniel Ramos, Raymond Veldhuis, Julian Fierrez, Luuk Spreeuwers, and Haiyun Xu, Regional fusion for highresolution palmprint recognition using spectral minutiae representation, IET Biometrics, Vol. 3, Issue. 2, pp. 94–100, 2014. Quan Wang, Kernel Principal Component Analysis and its Applications in Face Recognition and Active Shape Models, Computer Vision and Pattern Recognition, 2014. Saravanan Chandran and Satya Bhushan Verma, Touchless Palmprint Verification using Shock Filter, SIFT, I-RANSAC, and LPD, IOSR Journal of Computer Engineering, Volume 17, Issue 3, PP 01-08, 2015. Slobodan Ribari´c, and Marija Marˇceti´c, Personal Recognition Based on the Gabor Features of Color Palmprint Images, MIPRO May 21-25, 2012. T. Ojala, M. Pietikainen, and D. Harwood, Performance evaluation of texture measures with classification based on Kullback discrimination of distributions, Pattern Recognition, 1994. T. Ojala, M. Pietikäinen, D. Harwood, 1996. A comparative study of texture measures with classification based on featured distributions, Pattern Recognition, 29 51–59. T. Ojala, M. Pietikainen, T. Maenpaa, Multiresolution gray-scale and rotation invariant texture classification with local binary Satya Bhushan Verma is born in Barabanki, Uttar Pradesh, India, on 2012-1986. He completed B.Tech in Computer Science & Engineering from U.P. Technical University Lucknow, and completed M.Tech in Computer Science from MM University Mullana, Ambala, Haryana, India. Currently he is pursuing Ph.D. at the National Institute of Technology Durgapur, West Bengal, India. He has filed 1 patent, published 6 papers in International peer reviewed Journals/conference. Satya Bhushan Verma is member of IAENG (International Association of Engineers) Hong Kong. Saravanan Chandran is born in Tiruchirappalli, Tamilnadu, India, on 0101-1973. He has completed Ph.D. from Department of Computer Applications, National Institute of Technology, Tiruchirappalli, Tamilnadu, India, entitled “Analysis and Modelling of Grey-Scale Image Compression” in the year 2009. He worked as Programming Assistant at Bharathidasan University, Tiruchirappalli, Tamilnadu, India, from 1996 to 2000. He worked as Computer Programmer at National Institute of Technology, Tiruchirappalli, Tamilnadu, India from 2000 to 2007. He is working as Assistant Professor, Department of Computer Science and Engineering, National Institute of Technology, Durgapur, West Bengal, India, from 2007 to till date. He has filed 2 patents, published 32 papers in International peer reviewed Journals, 32 papers in national / international conference, 3 e-books, and chapters in two books. He is Senior Member of IEEE, Professional member of ACM, Life member of CSI and ISTE, Senior member of IACSIT, Singapore, member of IAENG, Hongkong. He is also serving as Editorial Board Member and Reviewer for several peer reviewed international journals. 187 https://sites.google.com/site/ijcsis/ ISSN 1947-5500