Increasing security with correlation-based fingerprint matching

The new generation of high resolution and touch-less fingerprint sensors can provide high resolution features, also known as level 3 features, in addition to macroscopic and microscopic features. Level 3 features matching will increase the system security to the governmental and Police levels. Unfortunately, traditional fingerprint matching techniques are not adapted to this new scenario and cannot exploit and match all the information the new fingerprints have confined inside them. In this paper a fingerprint matching algorithm that makes use of this new and relevant information is presented. The matching algorithm proposed is based in cross-correlation computation. Correlation uses the gray level information of the fingerprint image and can take into account all dimensional attributes of a fingerprint. In order to reduce the computational effort a coarse alignment step is performed beforehand. This alignment Is based in the cross-correlation of the orientation fields of the fingerprints. Region selection is performed to detect relevant information inside the fingerprint. For low resolution fingerprints the selection is based in the overlapping and the coherences of the orientation of both fingerprints. For high resolution fingerprints standard deviation of wavelet coefficients is also considered in the selection criterion. Experimental results demonstrate that this algorithm provides good results for both high and low resolution images.