Model based algorithm for singular point detection from fingerprint images

Abstract

The detection of singular points (SP) is an important step for automatic fingerprint identification system (APIS). In this paper, a model-based algorithm for singular points detection is proposed. Based on a model, which can describe the orientation fields around the singular point, a Hough transform is taken to compute the coefficients of the model, i.e., the exact position of the singular point. Since more information about singular point is used, the proposed algorithm is supposed to have a better performance than traditional Poincare index based algorithm. At the same time, the voting value of the Hough transform indicates the likelihood of the detected singular point, which can be used to eliminate spurious singular points. The experiments are taken to illustrate the algorithm's effectiveness.