A multiresolution method for singular points detection in fingerprint images

Abstract

The detection of the singular points (cores and deltas) is the most important task of the fingerprint image classification operation. We propose a new method of detection and localization of singular points in a fingerprint image. Dealing with a multiresolution representation of both the orientation field and the certainty level, the proposed method achieves a quite high precision in the localization of the singular points. A 2/spl times/2 pixels block at a certain resolution level is classified as ordinary, core or delta region based on the existent relationships among the directions exhibited by its four pixels. An index equivalent with the Poincare index is proposed here to perform this classification.