Fingerprint Verification Using Perturbation Method

This paper describes a new, powerful technique of fingerprint verification based on a perturbation method. The proposed method consists of four parts. The first part performs local FFT band-pass filtering to enhance the cyclic ridge structure in respective local areas. The second part is optimal block-wise shift for preliminary matching. Then, the third part is application of GAT correlation to realize affine-invariant shape matching. Finally, the fourth part is detail matching by perturbation. The key ideas of our perturbation method are in three ways: extraction of core points from enrolled fingerprint images, setting local windows around the core point, and asynchronous perturbation of local windows for optimal detail matching between input and each enrolled fingerprint images. How to design the size of local windows, the range and direction of perturbation, and the matching criteria is crucial to the success of the proposed method. Experimental results using the public FVC2000 fingerprint image database demonstrate a sufficiently low equal error rate (EER) of 5.55% for false rejection and false acceptance comparable to those obtained by competing works.