Fast and Robust Projective Matching for Fingerprints Using Geometric Hashing

Fingerprint matching is the most important module in automatic person recognition using fingerprints. We model the nonlinear distortions and noise obtained during the fingerprint capture process as transformations in the projective domain. Matching of the fingerprints involves computing the homography matrix for the projective transformation, mapping of the minutia points by this homography and finally computing the points that match each other within a pre-determined threshold. We perform a fast match using a Geometric Hashing-based algorithm which matches the points in polynomial time. Preliminary experiments with a sample representative database show promising results.