Face Spoofing Detection using Binary Gradient Orientation Pattern with Deep Neural Network
M. Parisa Beham, S. Md. Mansoor Roomi, H. Jebina, M. Kavitha · 2017
One of the challenging tasks in biometric authentication system is face spoofing attack. This paper presents an efficient and compact binary pattern based on gradient orientation with deep neural network for face liveliness detection. To increase the discrimination power of the proposed feature, multi scale retinex (MSR) normalization technique has been applied on the raw face image. Gradients are computed from several orientations to obtain the essential gradient oriented binary patterns (BGOP) in the normalized face. Finally, we implement deep neural network to learn BGOP features of high discriminative ability in a supervised manner. Combining with the deep network, the proposed BGOP feature achieved good error detection rate on standard datasets.