Robust Core-Point-ROI Based Fingerprint Identification Using a Sparse Classifier

We address the problem of automated fingerprints-based person identification from poor-quality fingerprints. Our solution includes the definition of a region of interest centered over the fingerprint at a reference point. From this region of interest a feature vector is computed that is invariant to some geometrical transforms but also to point transforms of the gray levels in the region of interest. This feature vector is then classified by means of a sparse classifier. We successfully test our algorithms on a publicly available fingerprints database and show that they are robust to a set of issues afflicting current fingerprint-identification systems in the case of poor-quality fingerprints.