Prediction Compression and Communication Based on Shared a priori Knowledge

摘要

Due to the importance of image communication, many approaches have been proposed to coding of the image sequence. In order to attain a low bit rate, some kinds of shared knowledge are needed. Inspired by this idea, a coding method based on the shared a priori knowledge and the block prediction is presented in this paper. Our method aims at the specific application field with relatively fixed roles and scenes on such case as newcasters and the environment they stay. By making use of a very small image base shared, any new input image which is an intermedia image of those images in the image base in sense of three dimensions can be predicted quite well. Such image base may be either built up by learning or stored in both ends in advance. The data transmitted mainly contriute themselves to those changed by facial expressions. Consequently, a high compression ratio can be reached.