An Optimized Approach for Fingerprint Binarization

This paper presents a novel approach for fingerprint binarization. The motivation is to simulate the human exercise that distinguishes ridges from valleys in uniform areas. An algorithm for performing such a region generation on fingerprint images is presented, and a data structure for saving generated regions is developed. Thresholding and binarization is then performed individually on each generated region using traditional binarization methods. The presented approach is particularly useful for processing fingerprint images captured with non-uniform inking or light. Experimental results show that this algorithm can achieve promising results compared with traditional binarization approaches.