A System Of Cartoon Facial Animation Based On Video

Video-based cartoon animation is one of the newly approaches in computer aided cartoon animation. And it is also one of the focus areas in recently researches. In this approach, computer vision techniques and image processing techniques are used to get useful information, such as color, profile, motion and so on, from the source video and then accomplish the cartoon animation with this information.In this paper, we will use the approach of video-based cartoon animation to assisted facial cartoon animation. In our paper, we designed a system of facial cartoon animation based on video, and we want to use the data which extracted from the video to drive the cartoon model, so we can simulate the expression from the source video.In order to achieve the above goal, we divide the facial animation into three situations: 1) non-expression when face rotates; 2) non-rotation when expression changes; 3) expression changes with head rotation. In view of the first kind of situation, the most important problem is to estimate the orientation of the face. So in this paper, we studied the approach of how to estimate the pose of face, and also we used an approach which based on camera calibration to estimate the orientation of face. Regarding second, third kind of situation, the difficulty is how to use the expression data which are extracted from original video to drive the cartoon model. This involves how to obtain the facial animation parameters. In our paper, we mainly introduced the solution about how to extract animation parameters.In order to obtain the facial motion data from the source video, we must first carry out the detection and tracking of facial feature. For this reason, we study the algorithm of facial feature detection which based on Active Shape Model, and the Lucas-Kanade optical flow tracking algorithm. And finally, we can use the motion data to drive the cartoon model to accomplish the facial cartoon animation.