Facial Contour Extraction Based on Level Set and Grey Prediction in Sequence Images

To improve the real-time performance of the level set algorithm for facial contour extraction in dynamic image sequence, a new facial contour extraction method is purposed in this paper which combines the level set and GM (1,1) model. With this method, firstly, the rough facial contour of a moving body is detected while it enters into the camera scope using the human motion information with the human skin-color model. Then the improved level set algorithm is used to accurately extract the facial contour. According to the integrity of facial contour motion, the centroid displacement of facial contour is predicted with GM(1,1) model so that the motion law of the facial contour can be estimated, which then can be taken as the iteration basis for the level set algorithm. In the same way, the extracted centroid of facial contour is taken as the prediction basis of the next frame GM(1,1) model. Experimental results show that this method can extract the motion facial contour with strong real-time performance and good robustness.