Fingerprint Matching using Enhanced Shape Context

Abstract

Shape context, a robust descriptor for point pattern matching, is applied in fingerprint matching by enhancing with minutiae type and angle details. A modified matching cost between shape contexts, by including the application specific contextual information, improves the accuracy of matching when compared with the original definition. To reduce computation for practical use, a simple pre-processing step termed elliptical region filtering is applied in removing spurious minutiae prior to matching. Empirical experiments conducted on a database of fingerprint images confirmed the improvements in accuracy and speed attained by the proposed method.