Fingerprint Recognition Based on Multi-Resolution Histogram of Gradient Descriptors

Abstract

This paper proposes a method to extract the fingerprint feature by using multi-resolution Histogram of Oriented Gradient representation. In this work, fingerprint is first enhanced for better ridge appearance. Next, Histogram of Oriented Gradient (HOG) is applied to model ridge valley structure as the occurrence of gradient information into a histogram bin size to obtain the fingerprint descriptor. Specifically, Multi-resolution fingerprint representation with HOG descriptor is used to isolate and analyse the fingerprint ridge structures in different resolution for better recognition performance. Experimental analysis shows that the proposed method is feasible in both performance accuracy and computational time as compared to conventional methods.