Efficient structure of video coders with motion-compensated fine-granularity scalability

The paper describes a two-loop coder structure suitable for mixed spatial and temporal scalability combined with fine-granularity. The system exploits independent motion estimation and compensation for each loop. The additional bitstream related to the increased number of motion vectors is compensated by more precise motion compensated performed independently at each level of resolution. The measures to limit the drift are also included. Drift is restricted to one loop only. Moreover, within a video sequence drift is periodically set to zero. The structure is suitable for encoding of video sequences within the framework of hybrid video coders like MPEG-2/4, H.263 and H.26L. The experimental results are presented in order to prove small scalability overhead in the bitrate, which has been tested for a wide range of bitrates as well as for different reference coders (H.263, MPEG-2).