International Journal of Computer Applications (0975 – 8887) Volume 53– No.10, September 2012 11 Fingerprint Identification System using Tree based Matching Subrat Kumar Sahu Department of CSE, National Institute of Science and Technology, Berhampur, Odisha, India Sruti Sahani Department of CSE, National Institute of Science and Technology, Berhampur, Odisha, India Pradeep Kumar Jena Department of CSE, National Institute of Science and Technology, Berhampur, Odisha, India Subhagata Chattopadhyay (corresponding author) Department of CSE, Bankura Unnayani Institute of Engineering, Bankura-722146, West Bengal, India ABSTRACT With the increasing focus on the automatic personal identification applications, biometrics specifically fingerprint identification is the most reliable, secure and widely accepted technique. The automatic fingerprint identification systems have two important steps, such as fingerprint (a) image enhancement and (b) minutiae matching. In this paper, we develop a fingerprint image enhancement as well as matching algorithm based on directional curvature technique (DCT) of local ridges and a modified Tree based matching approach. In the preprocessing stage, the Fingerprint is De-noised, Binarised, Thinned and the approximate core points are calculated by DCT algorithm. The Minutiae points are extracted by template filtering over the image. Identifying all the minutiae accurately as well as rejecting false minutiae is another issue, addressed in this paper. The Minutiae Matching Score is determined using a modified Tree Matching algorithm with assigned probability value with its level priority. The study reveals that the proposed modified Tree Matching algorithm has better matching percentage for different fingerprints as well as low quality fingerprint image compared to the existing algorithms. Keywords-Biometrics;Fingerprint identification; Directional curvature technique; binarisation; Tree matching algorithm; Minutae matching score. 1. INTRODUCTION Fingerprints are impressions of patterns left by friction ridges of the fingers’ skin (Gaensslen, 1991). The uniqueness of such patterns are recognized and used as one’s personal identification. It has the ability to confirm the identity of suspects and also to find the identity of unknown persons from prints left on the surface of the substrate. Thus the law enforcement and investigating agencies are the ones who have made the technology more reliable and secure for accurate identification. Human fingerprints can provide a sophisticated method of personal identification for various other applications. These include security or access control systems, banking and credit systems and forensic systems, registration process of individual. However, a shortcoming with many of the existing recognition techniques is that they all require a standard input image. Many sophisticated fingerprint enhancement algorithms exist which act differently for change in input image quality (Sherlock et al., 1994; O'Gorman et al., 1989; Methre, 1993). However, these techniques are suited for crime investigation where processing time does not matter a lot. (Coetzee and Botha, 1993) have investigated this problem up to a remarkable level. They proposed a recursive method following binarization and smoothing pre-processing, followed by feature extraction using a Fourier wedge-ring detector for detecting minutia. In general practice, a fingerprint quality may be degraded by noise due to impression, skin condition, scanner device etc., during image acquisition. To address such an important issue, fingerprint enhancement techniques are used to recover the structure of ridges and valleys from the noisy image by reducing quantifiable noise from image. Most of the fingerprint enhancement algorithms are based on the estimation of the orientation field (Huang, 1993; Jain et al., 1997; Lin et al., 1998). The most identifiable model for identification system is the minutiae-coordinate matching model. The two most prominent structures present on the fingerprint are ridge terminations and ridge bifurcations, which are usually called minutiae. In the existing system many schemes make use of local feature points i.e. minutiae based fingerprint matching systems (O'Gorman et al., 1989; Methre 1993; Coetzee and Botha 1993; and Huang, 1993) or exclusively global feature patterns (Methre 1993; Jain et al., 1997; Lin et al., 1998). A number of matching algorithms have been proposed in this context, such as the relaxation approach (Ranade and Rosenfeld, 1980), the fast algorithm based on 2-D clusters (Chang et al., 1997), the triangular matching and dynamic time warping approach (Mikl_os and Kov_acs-Vajna, 2000) and the local and global structure matching approach (Jiang and Yau, 2000). Jain et al. (1997) proposed an algorithm for point pattern matching in fingerprint recognition by its polar coordinate system and matched by a string matching algorithm. It is interesting to note that some key problems come across the existing recognition systems, which include (i) presence of noise in the fingerprint and (ii) false acceptance of the fingerprint due to degraded quality of the image. Thus a substantial amount of research reported in the field of fingerprint identification, especially devoted to rectify the image enhancement techniques, thinning and reliable method for feature extraction from the image. The matching stage uses the position and orientation of the minutiae features. As a result the reliability and exactness of feature extraction is crucial in the performance of fingerprint matching. In this paper we have concentrated on the thinning and matching algorithm for the identification process. The proposed algorithm forms a spanning tree by taking the identified minutiae points from the thinned image. The detail approach of this process is described in later part of the paper. The thinning process uses a modified approach of iterative Rotation Invariant Thinning Algorithm (RITA) proposed by (Patil et al., 2005). This process itself ensures the correctly identifying the minutiae point. Because, in any recognition system, thinning process takes an important as well as difficult part for exactly preparing the input image for the feature extraction step. A comprehensive survey of thinning algorithms is described by (Lam et al., 2000). In this modification, we have done the parallel processing of the