Fingerprint Liveness Detection Using Histogram of Oriented Gradient Based Texture Feature

Fingerprint authentication systems are widely used in civilian applications and governments. However, current problem is that fingerprints are easy to be imitated by artificial materials. Thus, spoof fingerprint detection has become increasingly significant. To solve the problem, a anti-spoofing mechanism, called FLD (fingerprint liveness detection), is used to discriminate spoof fingerprints from authentic fingerprints. In this paper, a new detection method using the histogram of oriented gradient with gamma correction is proposed. Firstly, gamma correction operation is used to eliminate the effects of poor quality fingerprints due to bad light and high brightness. Next, the sizes of captured images using different sensors are various, so images are zoomed to same scales through a bi-linear difference method. Then, features descriptors are described by the calculation of histogram of oriented gradient of pixels. Normalization operation is performed to remove the influence of abnormal features. Finally, the features representations are fed into SVM classifier with RBF. Experimental results on LivDet 2013 indicate that our method can yield a better classification performance compared with other methods.