Detection of falsified face using motion information

This study proposes a system for detecting falsified faces such as photographs of faces and masks by incorporating motion information in face verification and recognition systems. To enhance the robustness of the falsification detection system, a multi-frame optical flow method, which generates significant differences between the optical fields of true faces and that of falsified faces, is adopted. After the optical flow fields are obtained, two classification schemes are applied in the detection system to distinguish true faces and falsified faces, and their results compared. Under the specified motions of true faces and falsified faces, the proposed method can detect falsification with high accuracy. Even when the motion is arbitrary, the proposed system provides satisfying results.