Monocular camera-based face liveness detection by combining eyeblink and scene context

Abstract

This paper presents a face liveness detection system against spoofing with photographs, videos, and 3D models of a valid user in a face recognition system. Anti-spoofing clues inside and outside a face are both exploited in our system. The inside-face clues of spontaneous eyeblinks are employed for anti-spoofing of photographs and 3D models. The outside-face clues of scene context are used for anti-spoofing of video replays. The system does not need user collaborations, i.e. it runs in a non-intrusive manner. In our system, the eyeblink detection is formulated as an inference problem of an undirected conditional graphical framework which models contextual dependencies in blink image sequences. The scene context clue is found by comparing the difference of regions of interest between the reference scene image and the input one, which is based on the similarity computed by local binary pattern descriptors on a series of fiducial points extracted in scale space. Extensive experiments are carried out to show the effectiveness of our system.