Fingerprint Matching Using Distortion Tolerant Filters
Craig I. Watson, D P. Casasent
The purpose of this paper is to discuss the use of distortion tolerant filters to improve the performance of fingerprint correlation matching. Previous work has shown that correlation matching of fingerprints is susceptible to larger false alarm rates than more traditional minutiae based methods. The elasticity of the finger-print creates enough distortion to produce poor results in correlation matching, be-cause the correlation matcher uses the global ridge structure of the fingerprint image. A way to improve correlation matcher performance is to create filters that combine several elastic distorted versions of the fingerprint images. In this work three different distortion tolerant filters are tested: average, synthetic discriminate function (SDF), and minimum average noise and correlation plane energy (MINACE).