Color image sequence coding with variable block size motion compensation

In this paper, a new color image sequence coding method based on variable block size motion compensation is proposed. A multiresolution tree structure is used to organize the blocks with various sizes. At the bottom level, the quadtree is used to code the resultant motion compensated prediction error. The spatiotemporal characteristics of the human eyes are embedded into the tree structure to achieve high compression. For transmission of color information in the scenery, the color pixel is first transformed into Y-U-V color space, and then a color quadtree structure is constructed to transmit the full color information. Massive experiments have been done with the CLAIRE image sequence. Results show quite promising.<>