Face antispoofing based on frame difference and multilevel representation

摘要: Due to advances in technology, today’s biometric systems become vulnerable spoof attacks made by fake faces. These occur when an intruder attempts fool established face-based recognition system presenting a face (e.g., print photo or replay attacks) front of the camera instead intruder’s genuine face. For this purpose, antispoofing has hot topic analysis literature, where several applications with task have emerged recently. We propose solution for distinguishing between real faces and ones. Our approach is based on extracting features from difference successive frames individual frames. also used multilevel representation that divides frame into multiple multiblocks. Different texture descriptors (local binary patterns, local phase quantization, binarized statistical image features) then been applied each block. After feature extraction step, Fisher score sort ascending order according associated weights. Finally, support vector machine differentiate tested our three publicly available databases: CASIA Face Antispoofing database, Replay-Attack MSU Mobile Spoofing database. The proposed outperforms other state-of-the-art methods different media quality metrics.