Distributed Hypergraph Laplacian for Face Anti -spoofing with Monocular Images

Face anti-spoofing is an important part of face recognition. However, anti-spoofing with monocular images is still challenging due to the complex appearance of facial images. Therefore, current methods require a large number of training data. To improve data utilization and provide privacy preservation, we propose a novel face anti -spoofing method based on distributed learning in this paper. The key contribution is called Distributed Hypergraph Laplacian (DHL). It learns the semantic relationships among training data with manifold learning in a distributed way. In this way, images can be represented by the eigen decomposition of the manifold matrix and the original data will not be observed. Besides, hypergraph Laplacian is introduced to further improve the robustness of manifold learning. Experiments are conducted on two popular benchmark datasets MSSPOOF and CASIA-SURF. The results show the effectiveness of DHL.