Anisotropic Smoothing for Illumination Invariant Face Anti-spoofing

Abstract

Presently, in various biometric-based access systems face recognition has been used. But few of the systems have the facility to discriminate real and spoof faces. Moreover, plenty of researchers on face live detection primarily concentrated on intrusive approaches, which are non-user approachable in nature. An innovative non-intrusive face spoofing detection technique based on HOG and ELTP descriptors is proposed in this paper. More precisely, this approach adopts anisotropic smoothing for normalizing the illumination of a given image. The effectiveness of the proposed method has been evaluated by different types of experiments on the NUAA dataset.