Improvement of fingerprint matching by describing the minutiae neighborhood using a set of Quaternion Disc-Harmonic Moments

Abstract

we propose a hybrid method that integrates the minutiae and their local neighborhood information. This contribution can be seen as a validation step of the minutiae triplets matching performed in a recent algorithm: M3gl. We aim to improve the result of the minutiae triplets matching stage. The characterization of the neighborhood of each triplet is established by a unique feature vector. This descriptor is based on a novel set of moments: the Quaternion Disc-Harmonic Moments (QDHMs). The QDHMs were proposed as an extension of the ordinary disc-harmonic moments using the algebra of quaternions. This representation is able to distinguish minutiae having different neighborhood information. We carried several experiments using the database FVC2002 data and the results underline a significant improvement of the classical M3gl.