An Efficient Graph-Based Algorithm for Fingerprint Representation and Matching

Abstract

Abstr act.This paper presents a novel graph-based algorithm for robust fingerprint recognition, which efficiently improves the approach ``K-plet and CBFS'' proposed by Chikkerur et al at ICB'06.To cope with the crucial problems of nonlinear distortion and partial overlap, a new feature called supplementary associative point (SAP) is extracted to provide additional discriminative information for representation and matching.The local spatial relationship of the improved K-plets, which are constructed by minutiae and SAPs, is encoded as a directed graph to represent the fingerprint.Besides, an improved CBFS-based matching algorithm is also proposed.It utilizes the consistent matching property of minutia-SAP pairs, and possesses advantages such as fast pruning of imposter local matches and deep propagation of local matches in graph.Experiments on FVC database show that the proposed algorithm is strongly robust to nonlinear distortion and partial overlap, and outperforms the original algorithm with an average improvement of 24.17% on EER.