An improved iris recognition system based on Gabor filters

Abstract

An improved iris recognition system based on even symmetry Gabor wavelet is proposed in this paper. In iris image preprocess, an effective segmentation and eyelids occlusion detection approach are introduced. During recognition, only the imaginary part of complex 2-D Gabor filter is used in coding for shorter code length but without affecting the recognition performance. A simple effective filter parameters selection scheme is proposed at the same time. Experimental results on CASIA1.0 and ICE 2005 databases clearly demonstrate that our improved methods can achieve highly accurate iris recognition with a short code length.