Rotation-invariant Weber pattern and Gabor feature for fingerprint liveness detection

Fingerprint recognition systems are extensively deployed for the authentication in many applications. However, this kind of recognition systems may be spoofed by artificial fingerprints made from various materials. Thus, it is necessary to add a fingerprint liveness detection module to keep this kind of recognition systems on a good level of security. The fingerprint liveness detection (FLD) aims to judge whether a given fingerprint image is captured from a real finger or a spoof one. It is a typical two-class classification problem where the feature extraction is the key step. In this paper, we propose an effective feature extraction method for the FLD problem. The proposed features consist of two components, Weber local binary pattern (WLBP) and circularly symmetric Gabor feature (CSGF), analyzing the fingerprint images in both the spatial and frequency domains. The co-occurrence probabilities of the two components are calculated as the final features. The proposed features are utilized to train SVM classifiers separately on two databases in Fingerprint Liveness Detection Competition 2011 and 2013. Experimental results demonstrate the effectiveness of the proposed method.