Face Detection And Optical Flow Tracking In Video Sequences

Face detection and tracking technology is a key technology in face recognition, video retrieval, virtual gaming and graphics control, It is great significant to promote the further face tracking development, it has become a hot topic of computer vision research.In this paper, the difference image is calculated at each pixel in the face region mean square gradient matrix, Eigenvalue analysis is given to find the corners of the face region by the respond of corner. Because the corners are easily been tracked, so the optical flow theory is introduced to calculate the corners' optical flow between two frames, in order to achieve face tracking. If the corner's local area is a constant or no texture changes, the corner's optical flow can not be calculated, to solve this problem, color constant constraint is introduced in the optical flow calculation. Because the corners are not sensitive to light and noise, this algorithm improved the robustness of face tracking in light changes and face gesture. At the same time, optical flow algorithm is used for the face the track, rich facial motion information can be also got. Experimental results show that the face in video with gesture,facial expressions and light changes can be reliable tracked, it is an effective method of face tracking.