Image based fingerprint verification

Abstract

Human fingerprints have been considered as a unique signature certifying one's identity. Fingerprint verification is currently the most popular technique of biometric personal identification. As a biometric proof of identification, an alignment-based minutiae matching algorithm has been widely investigated, however, not many have considered fingerprint identification using image based verification. An image-based approach does not use minutiae features for fingerprint matching. This project investigates fingerprint-scanned image verification via an inverse fast Fourier transform after a thinning process. The technique may be applied directly to a gray-scale fingerprint image without pre-processing. This correlation coefficient approach is capable of finding correspondences between the input fingerprint image and the stored enrolled template with higher computational efficiency than the minutiae-based method. It has the ability of robust fingerprint verification subject to external conditions. In this paper, the design and implementation of a prototype of an automatic verification system is presented. The high matching rate achieved in this project suggests that an efficient small-scale fingerprint verification system is feasible.