Accurate and quasi-automatic lip tracking

摘要: Lip segmentation is an essential stage in many multimedia systems such as videoconferencing, lip reading, or low-bit-rate coding communication systems. In this paper, we propose accurate and robust quasi-automatic algorithm. First, the upper mouth boundary several characteristic points are detected first frame by using a new kind of active contour: "jumping snake." Unlike classic snakes, it can be initialized far from final edge adjustment its parameters easy intuitive. Then, to achieve segmentation, parametric model composed cubic curves. Its high flexibility enables contour extraction even challenging case very asymmetric mouth. Compared existing models, brings significant improvement accuracy realism. The following frames achieved interframe tracking keypoints parameters. However, show that, with usual algorithm, keypoints' positions become unreliable after few frames. We therefore process that hundreds Finally, mean errors our algorithm comparable manual points' selection errors.