Low Complexity Low Delay Fgs Coding with Adaptvie Reference

In this paper, a fine granularity SNR scalable video coding scheme is presented. Targeted for low delay applications requiring high coding efficiency, temporal prediction is incorporated into the FGS layer of closedloop P frames with the prediction formed adaptively from enhancement layer and base layer references. This generally requires motion compensation at both enhancement layer and base layer for decoding a FGS frame, which results in high computation complexity when multiple FGS layers are involved. The coding scheme presented in this paper can efficiently reduce the complexity in such cases. With this method, two motion compensations are sufficient for decoding a FGS frame regardless of which FGS layer is to be decoded, while preserving good coding efficiency.