Palmprint Recognition by Wavelet Transform with Competitive Index and PCA Deepti Tamrakar, Pritee Khanna Abstract—This manuscript presents, palmprint recognition by combining different texture extraction approaches with high accuracy. The Region of Interest (ROI) is decomposed into different frequency- time sub-bands by wavelet transform up-to two levels and only the approximate image of two levels is selected, which is known as Approximate Image ROI (AIROI). This AIROI has information of principal lines of the palm. The Competitive Index is used as the features of the palmprint, in which six Gabor filters of different orientations convolve with the palmprint image to extract the ori- entation information from the image. The winner-take-all strategy is used to select dominant orientation for each pixel, which is known as Competitive Index. Further, PCA is applied to select highly uncorrelated Competitive Index features, to reduce the dimensions of the feature vector, and to project the features on Eigen space. The similarity of two palmprints is measured by the Euclidean distance metrics. The algorithm is tested on Hong Kong PolyU palmprint database. Different AIROI of different wavelet filter families are also tested with the Competitive Index and PCA. AIROI of db7 wavelet filter achievs Equal Error Rate (EER) of 0.0152% and Genuine Acceptance Rate (GAR) of 99.67% on the palm database of Hong Kong PolyU. Keywords—DWT, EER, Euclidean Distance, Gabor filter, PCA, ROI. I. I NTRODUCTION Palmprint based personal recognition has become an active research topic in recent years. Compared with the other currently available biometric features, palmprints contain more distinctive information than fingerprints; palmprints acquisi- tion devices are much cheaper than iris devices; palmprints can build highly accurate biometrics system than face and voice. Palmprint has several advantages such as stable line features, low-resolution images, low cost acquisition device, very difficult to fake, and easy to user acceptation etc. Palm- print has features like texture, wrinkles, principal lines, ridges, and minutiae points that can be used for its representation [1]. Texture and palm lines are the most clearly observable palmprint features in low resolution (such as 100 dpi) images [1], and thus have attracted most research efforts. Therefore texture based approaches are adopted for palmprint recogni- tion. Typically, texture features are extracted by filtering the palmprint images using filters such as the Gabor filter [2], [3], [4], [6], [7], [8], [10], Guassian filter [9], Radon filter [11], and wavelet transform (WT) [12], [10], [13]. The common tasks in texture-based approaches are to extract palm line orientation and compare similarity between different images. The WT has been widely used for palmprint recognition in different forms. The local or global statistical features (energy, entropy) Pandit Dwarka Prasad Mishra Indian Institute of Information Technology, Design & Manufacturing Jabalpur. e-mail: pkhanna@iiitdmj.ac.in of the sub bands of multi resolution analysis are used for palmprint recognition [3], [4], [6], [10]. A different form of Gabor filter is widely used for palmprint recognition. Several coding methods have been developed using Gabor filter such as Competitive Code [8], [9], [11], [12], [5], Fusion Code[12], Binary Co-occurrence Vector[14] and Palm Code [10]. Coding based methods are seemed to be the most promising due to the strengths of small feature size and high recognition accuracy [8]. The orientation and phase information of palm lines are typically encoded as binary or integer numbers, which are robust to illumination variations [8], [10]. However, the number of features in the coding method is equal to the size of image and each feature is represented by binary code. Phase information of the palmprint gives the efficient results without binary coding. But problem is the dimension of the feature vector, which is very large and needs to be reduced. Several statistical subspace approaches exists for reduction of the size of feature vector without loss of any potential information. These approaches can find the uncorrelated features which can discriminate the different classes of the images more accurately. Some Subspace-based approaches of dimensional reduction and discrimination analysis are also used for the palmprint recognition such as Principal Component Analysis (PCA) [16], [17], [18], Fisher Discriminant Analysis [19], Linear Discrimination Analysis (LDA), Locality Preserving Projections (LPP) [20] and Independent Component Analysis (ICA). In literature, Subspace-based approaches are directly applied to palmprint images or combined with other tech- niques. Instead of taking original ROI, the proposed work is concentrated on the low resolution approximation ROI using DWT. Approximation ROI (AIROI) can reflect sufficient infor- mation of nearly all features of palmprints, so that palmprint can quite discriminate by approximation coefficients (without scalable detailed coefficients) [13]. In this works, Gabor filters based competitive index method is used which applied six Gabor filters of different orientation on the AIROI. The index of filtered image which has minimum value corresponding to each pixel is selected. The PCA is further applied to reduce the feature size and to select uncorrelated features. Rest of the paper is organized as follows; Section 2 explains the preprocessing for extraction of the region of the interest. Section 3 describes the extraction of AIROI by the DWT. Section 4 describes the Competitive index approach. Section 5 explains the Comp Index with PCA. Section 6 gives distance metrics for matching of the two palmprint images. Section 7 describes the experimental setup for proposed technique. Results of the proposed technique and its comparision with other existing methods of palmprint recognition are given in World Academy of Science, Engineering and Technology International Journal of Computer and Information Engineering Vol:5, No:12, 2011 1621 International Scholarly and Scientific Research & Innovation 5(12) 2011 ISNI:0000000091950263 Open Science Index, Computer and Information Engineering Vol:5, No:12, 2011 publications.waset.org/5291/pdf