Removal of False Minutiae based on Structural and Directional Attributes in Fingerprint Recognition

초록

The feature template of a fingerprint in automatic fingerprint recognition contains false minutiae due to noise during image acquisition and/or in-formation loss in pre-processing. This paper proposes a new method of remov-ing false minutiae in fingerprint feature templates. Smudgy regions in a fingerprint image largely produce false minutiae, where after thinning, the ridges need to be restructured. The proposed algorithm determines false minu-tiae among minutia candidates based on a series of structural characteristics and directional consistency of ridge flows in the thinned fingerprint image. The al-gorithm proposes a set of rules of restructuring ridges, where it utilizes the av-erage distance between the ridges of individual fingerprints in order to make it-self more adaptive and robust. The algorithm has been implemented and applied to the NIST Special DB 14. The results show remarkable reduction in false mi-nutia rate with the cost of negligible dropping in true minutia rate, which con-sequently increases the overall performance of an automatic fingerprint recognition system.