Local Correlation-based Fingerprint Matching

Most fingerprint matching systems are based on matching minutia points between two fingerprint images. Each minutia is represented by a fixed number of attributes such as the location, orientation, type and other local information. A hard decision is made on the match between a pair of minutiae based on the similarity of these attributes. In this paper, we present a minutiae matching algorithm that uses spatial correlation of regions around the minutiae to ascertain the quality of each minutia match. The proposed algorithm has two main advantages. Since the gray level values of the pixels around a minutia point retain most of the local information, spatial correlation provides an accurate measure of the similarity between minutia regions. Secondly, no hard decision is made on the correspondence between a minutia pair. Instead the quality of all the minutiae matches are accumulated to arrive at the final matching score between the template and query fingerprint impressions. Experiments on a database of 160 users (4 impressions per finger) indicate that the proposed algorithm serves well to complement the 2D dynamic programming based minutiae matching technique; a combination of these two methods can reduce the false non-match rate by approximately 3.5% at a false match rate of 0.1%.