Face liveness detection using color gradient features

Face liveness detection is widely used to detect face spoofing attacks. Most existing face liveness detection methods are mainly based on texture information of face images in the gray-scale space which ignored the chrominance information and other feature differences. In this paper, we introduced a novel modified color gradient feature into face liveness detection which used variant color roberts cross operator. This method can extract the color gradient information from the live faces or fake faces so that the proposed method has potential to achieve a better performance. Extensive experimental results on benchmark databases showed that the proposed variant color-gradient feature was very effective for face liveness detection.