Model-Based Image Coding & Reconstructing For Image Sequences

In the movement of economy and society, the demand for real-time video data transportation increased steadily. As a result, the conflict between the limited network bandwidth and the enormous video data became more and more intensely and people are more and more interested in the low / very low bit-rate video encoding. ISO and ITU-T have presented several low bit-rate video compression protocols such as ISO-MPEG4/MPEG7, H.264, etc from 1993 when ITU proposed the H.263 protocols. Recently the model-based coding technology draws much more attention because encoding video data under above protocols often can not meet the demand for the very low bit-rate (about 10 bps) transportation. Using the model-based coding technology, first, a model is constructed based on anterior information of the image, then the model parameters are obtained and encoded. Because the encoded object is parameter not image, very high compression rate can be achieved with the model-based coding technology.In this paper, the elements of model-based coding technology for image sequences are presented first, then several conventional algorithms for model-based coding are discussed in detail and a simple and effective method for determining the positions of feature points which should be known before 3-D motion and structure estimation is proposed. In order to simplify the calculation during model-based coding, a new 3-D frame model for face and a reconstructing algorithm based on it are presented in chapter 4. Also the photometric effects are considered and the simulation results are given out. In the final part of this paper, future work is discussed.