Fingerprint Center Point Location Using Directional Field Y.H. Chan and S.A.R. Abu-Bakar Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia. {yhuichan@yahoo.com and syed@fke.utm.my } Abstract This paper presents a reliable fingerprint center point (CP) location algorithm for the alignment of fingerprints to construct a shift invariant fingerprint recognition system. The proposed algorithm is based on Alteration Tracking (AT) and CP estimation (CPE). AT is proposed to extract a track that records the transition from one quantized direction to another. CPE is aimed to find the bending point with highest transition of direction from the transition track. This algorithm is tested against fingerprints captured from SAGEM MSO100 optical scanner and the second database from University of Bologna. Experimental result shows that the proposed algorithm is capable of reliably locating fingerprint CP. 1. Introduction Automatic fingerprint recognition system often faces the problem of fingerprint translation, rotation and distortion during fingerprint acquisition. Two fingerprints from the same individual may suffer from these undesired situations. Hence, CP is used as a pre-processing stage for fingerprints alignment to construct a shift invariant system. Most of the available algorithm for locating a CP is by finding a core point from the directional field based on Poincaré index, such as [1] and [2]. It is due to the fact that Poincaré index is able to locate singular points, thus problem arises for arch fingerprints. Reference point location algorithm presented in [3] calculates the difference of sine component for orientation field between adjacent regions. [4] presents a simplified method for calculating fingerprint orientation by finding approximate gradients of the image, an intersection of a near zero transition for both approximate gradients in x- and y-axis is obtained from its neighbourhood and its singularity type is determined using derivatives from the two gradient vectors to calculate the angle between them. In this work, we define CP by using the definition of reference point proposed in [3]. The CP is the point of maximum curvature of the concave ridges in a fingerprint image [3]. The proposed algorithm in this work is to have a quick access to locate CPs for fingerprints of all classes, i.e. whorl, right, left loop and arch from fingerprint directional field. One CP is estimated for one fingerprint. This paper is organized as follows: Section 2 presents the directional image construction algorithm and grayscale opening operation. Section 3 illustrates AT algorithm. The CPE is explained in Section 4. Section 5 shows the performance of the proposed algorithm and finally conclusions are drawn in Section 6. 2. Fingerprint Directional Field Based on the work by [5], an input fingerprint image undergoes wavelet transform (WT) using Haar wavelet coefficients ] [ 2 1 2 1 for high pass filter and ] [ 2 1 2 1 for low pass filter. The second stage WT is done by using expanded coefficients for edge detection similar to Prewitt gradient filter. The expanded coefficients for high pass filter and low pass filter are respectively ] 0 [ 2 1 2 1 and ] 3 1 3 1 3 1 [ . Ridge orientation estimation is ) , ( ) , ( tan 2 2 1 ) , ( c r W c r W hori vert c r where W is vertical detail of 2nd stage WT ) , ( 2 c r vert W is horizontal detail of 2nd stage WT ) , ( 2 c r hori The directional image is quantized into 8 directions, smoothed using smoothing filter of size 7 7 and then down sampled with a factor of 2. The input fingerprint is N N and the obtained directional image is 8 8 N N . Grayscale opening operation is performed over the directional field to smooth the transition edge between grayscale value 144 representing +11.25° and grayscale value 108 representing -11.25°, and eliminating grayscale values which represent higher value of degree within region of lower value of degree. Proceedings of the Third International Conference on Image and Graphics (ICIG’04) 0-7695-2244-0/04 $20.00 © 2004 IEEE Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on March 25,2010 at 04:09:43 EDT from IEEE Xplore. Restrictions apply.