Context based face spoofing detection using active near-infrared images

摘要: In this paper, with the help of controllable active near-infrared (NIR) lights, we construct differential (NIRD) images. Based on reflection model, NIRD image is believed to contain lighting difference between images and without NIR lights. Two main characteristics based are exploited conduct spoofing detection. Firstly, there exist obviously media around faces in most conditions, which reflect incident lights almost same way as face areas do. We analyze pixel consistency non-face employ context clues distinguish Then, feature, extracted only from areas, utilized detect attacks deliberately cropped medium. Merging two features, present a detection system. several experiments self collected datasets different media, demonstrate excellent results robustness proposed method.