Support vector method for minutiae detection in fingerprint images

Fingerprint minutiae (i.e. ridge endings and ridge bifurcations) form a pattern that is unique to each fingerprint. Almost all automatic fingerprint comparison systems rely on minutiae matching, and hence the minutiae extraction from fingerprint images highly influences the performance of every such system. A minutiae extraction method based on support vector machines is proposed here. The method does not require a ridge thinning processing step in contrast with most of the previously proposed methods of minutiae detection and classification. Because of this, the number of spurious minutiae detected is maintained low, such that a subsequent processing step of spurious minutiae elimination becomes unnecessary.