High resolution fingerprint recognition using pore and edge descriptors

Pores have been used for fingerprint recognition for many years. Introducing pores into fingerprint image comparison can improve the recognition accuracy. This paper proposes a novel method to compare pores on high resolution fingerprint images. The method consists of three steps. In the first step, pores on different images are clustered using a Root-SIFT descriptor. Then the pores in the same class will be compared using a rotational invariant edge descriptor and Hough transform. The descriptor consists of three parts: the length of the edge, the angles between the edge and the orientations of pores, and the angle between the orientations of pores. In the final step, one-to-one correspondences of pores are established based on the matched edges, and then refined using the spatial relations of the pores. To this end, a graph based refinement algorithm is designed. Experimental results on two high resolution fingerprint image databases show that the proposed algorithm is more accurate than other state-of-the-art pore comparison algorithms.