Very low bit-rate video coding with block partitioning and adaptive selection of two time-differential frame memories

Recently several studies on very low bit rate video coding have been reported. One of the major goals of the studies is to improve coding performance, which gives better subjective and objective quality than conventional coding methods at the same bit rate. As the shape and structure of an object in a picture are arbitrary, the performance of traditional coding with block-based motion compensation (MC) is not satisfactory. In this paper, we present advanced MC schemes for very low bit-rate video coding. Major features of the proposed MC are block-partitioning prediction and utilization of two time-differential reference frames, This coding scheme improves image quality around objects' boundaries and consequently reduces prediction errors. It also works well in the case of object occlusions. The combination of the proposed MC and discrete cosine transformation (DCT) will show better performance in several test sequences than full-spec H.263.