Robust Multiple Minutiae Partitions for Fingerprint Authentication

Nonlinear distortion of fingerprint images still degrades system performance especially the fingerprint images generated by 1-line swipe sensors. In this paper, a robust fingerprint-matching algorithm based on multiple minutiae partitions (MMP) is presented to match general distorted fingerprints. Minutiae in an input and a template fingerprint are globally partitioned into pairs of multiple neighbors, each of which consist of similar local minutia topology and features. The minutia similarity is additionally evaluated by investigating ridge shape and the geometric similarities of the corresponding neighbor minutiae. The global similarity score is calculated with the consolidation of the derived multiple partitions, which should be subject to a reasonable condition of different implicit alignments between pairs of partitions. The proposed algorithm is evaluated using the database captured from a line sensor, and results show its wide coverage for vertical distorted fingerprints greatly improves the matching precision when compared to conventional approaches.