Fingerprint Matching Based on Directional Image Feature in Polar Coordinate System

Abstract

This paper presents a new fingerprint feature extraction and alignment method based on a directional image representation in polar coordinate system. First, the proposed method establishes a region of interest (ROI) for feature extraction using the reference point information. The ROI is then converted from a Cartesian coordinate system to a polar coordinate system to facilitate the following feature extraction and rotational alignment processes. In the proposed method, standard deviation value of each directional subband block is exploited as the fingerprint feature, and the directional subbands are obtained using a directional filter bank (DFB). Input feature vectors, in which various rotations are considered, are extracted by cyclically shifting the decomposed subband outputs and recalculating the directional feature value of each block, and these input feature vectors are matched with the enrolled single template feature vector. Rotational alignment is achieved by finding the minimum Euclidean distance. Experimental results demonstrated the effectiveness of the proposed method in feature extraction and alignment, along with a comparable verification accuracy to that of other leading techniques.