Face spoofing detection using dynamic texture

摘要: While there is a significant number of works addressing e.g. pose and illumination variation problems in face recognition, the vulnerabilities to spoofing attacks were mostly unexplored until very recently when an increasing attention started be paid this threat. A attack occurs person tries masquerade as someone else by wearing mask gain illegitimate access advantages. This work provides first investigation research literature on use dynamic texture for detection. Unlike masks 3D head models, real faces are indeed non-rigid objects with contractions facial muscles which result temporally deformed features such eye lids lips. Our key idea learn structure dynamics micro-textures that characterise only but not fake ones. Hence, we introduce novel appealing approach detection using spatiotemporal (dynamic texture) extensions highly popular local binary pattern approach. We experiment two publicly available databases consisting several different natures under varying conditions imaging qualities. The experiments show excellent results beyond state-of-the-art.