Model-Based Video Coding — Some Challenging Issues

Most of existing image coding techniques fall into the category of so-called waveform based methods with limited compression factors. In recent years, a new video compression approach using model-based coding techniques arises as more and more researchers seek for very low bit-rate video compression[1]–[4] The main idea of this approach is to use 3D models of moving objects and derive the motion parameters associated with the model structure, then the original video sequence could be approximately reconstructed using the 3D models and the motion parameters. At present, in most research activities on model-based coding systems, the dominant parts in the video scenes are human faces talking and showing expressions, for example, in videophone and teleconferencing. Therefore, three research topics — the face model, facial motion analysis and facial expression synthesis, are of paramount importance and many researchers in both computer vision and computer graphics have contributed to these themes significantly.