Fingerprint Orientation Modeling Using Symmetric Filters

Abstract

Accurate fingerprint orientation is a prerequisite in fingerprint based recognition system. This paper proposes an algorithm for modeling the fingerprint orientation field by using a model based algorithm based on the weighted Legendre basis. Weights required in the modeling are obtained by using symmetric filters, such that: i) high weights should be assigned to the areas near singular points, ii) areas having uniform ridge-valley flow should be given high weights, and iii) areas containing bad quality due to dry/ wet fingerprints, scars, bruises or sensor condition should be given low weights. These conditions ensure accurate reconstruction of fingerprint orientation field for bad quality areas while preserving the true orientation field near singular points. The proposed algorithm has been evaluated on a publicly available database, FVC2004 DB1A. Experimental results reveal that it has better orientation field estimation compared to the various state of the art algorithms.