An Efficient Algorithm for Fingerprint Reference-Point Detection

Abstract

The reference-point played important roles in most algorithms of fingerprint recognition. It is widely used in fingerprint classification, and fingerprint matching. There are many methods proposed for reference point detection as Poincare index technique, geometry of region technique, direction curvature technique and orientation-consistency-based technique. However, the experimental results show that the present methods are less accurate on the fingerprint database which contains many poor-quality fingerprint images (the most accuracy of the present methods is 93% on the fingerprint database FVC2004, DB2, set A). This paper proposes a novel algorithm to locate fingerprint reference-point consistently and accurately for all types of fingerprints. The method computes the convex orientation consistency which describes both direction of curvature and how well the orientations over a neighborhood are consistent with the dominant orientation. Experimental results show the effectiveness and superiority of our proposed method (96.9 % on the FVC2004, DB2 set A).