The main objective of this paper is to develop a fine grained scalable video codec which can fully exploit temporal correlations within all layers, and thus offer performance closer to that of a single layer non-scalable system. To this end we propose the use of a combination of matching pursuits together with absolute value coding of the displaced frame difference information. The proposed codec is also shown to be highly robust to channel errors. This results from a combination of the use of fixed length code-words and a reduction of temporal error propagation. The use of fixed length codewords also facilitates the use of lower complexity transcoding operations to support more sophisticated forms of scalability
展开▼
