Definition and extraction of stable points from fingerprint images

摘要

This paper introduces a new idea to define and extract singular points (SPs, including cores and deltas) from fingerprint images by a zone Could-be-in theorem that describes harmonic relationships of a SP and its neighbors. According to the theorem, we can view SPs as stable points (attractors and rejectors by their rotations), estimate core's direction, and restore SP offset. The proposed method is validated in the following two hierarchical aspects, one is to test the reliability and accuracy of coarse SP detection by means of two indices (RI and AC), and the other is the accuracy of core's location and the stability of core's direction in terms of improved FingerCode matching. Experiments on partial FVC2004 data sets show that it works well and outperforms three other methods.