Research On Face Tracking Approach In Video Sequence Images

At present, with the rapid development of information technology and the widely use of intelligent Human-Computer Interaction system, the automatic face detection and tracking had already become the popular research topic in computer vision and pattern recognition domain. With development of video technique, face detection and face tracking have broader prospects for development.The thesis proposed a face detection method based on Gauss skin color model, lip color model and the Geometric relationship between the features on face. Our approach is for color images, we construct a skin color model in YCbCr color space to segmented skin color region. Then we detect lip in potential face region using lip color model and searches the eyes using geometry textures. Results of the study demonstrate that the proposed approach performs well in detecting vary pose in different illumination conditions and complicated backgrounds and can used for video sequence and static images.The thesis proposed a face tracking method using particle filter based on skin color model and facial contour. We use the multi-peak characteristics of particle filter combining color information and the face contour characteristics to track face. Comprised with the traditional single trail tracking method based on skin color, this method may solve partial occlusion problems, and many video sequences have similar skin color region in complicated backgrounds, to improve the tracking robustness.