Fingerprints Matching Based on Quadrangle

Abstract

A novel fingerprint matching algorithm based on quadrangles has been developed. To begin with, we extracted ridge lines from fingerprints, and three phases were performed to match the ridge lines between two fingerprints. In the first phase, an inaccurate set of matched line pairs were built by matching the local quadrangle set (LQS). In the second phase, the fake pairs were removed from the set by evaluating the correlation among the pairs. A decision for the two fingerprints coming from the same finger or not could be made based on the shrunken set in case there were enough pairs in the set, otherwise the set was extended by adding the possible matched pairs that failed matching in the first phase so that a more reliable decision could be made. Experiments using the FVC2002 (Fingerprint verification competition 2000) databases show that the proposed algorithm is effective.