An embedded wavelet-based quadtree interframe coding algorithm

Motion compensated prediction (MCP) error image has significant amplitudes mainly along the boundaries of moving objects, correspondingly its wavelet coefficients contain large energy in high frequency bands. This greatly reduces the coding efficiency of embedded zerotree wavelet (EZW) coding. In this paper, we first provide the correspondence between the significant amplitudes area of the MCP error image and the wavelet coefficient significant map, then encode the significant map efficiently using the quadtree approach, followed by bit-plane coding the significant coefficients to produce an embedded data stream. Experiment result shows that compared to EZW and ZTE, the proposed algorithm is more efficient and has better performance in very low bit rate applications