METHOD OF SYNTHESIZING A PREDICTED IMAGE, VIDEO CODING DEVICE AND VIDEO CODING METHOD

Two arbitrary kinds of block images selected from the LMC block image, GMC block image and G-LMC block image, which are respectively synthesized by the local motion compensation, global motion compensation and global-local motion compensation, are averaged to produce an average block image, any one block image is selected from those LMC block image, GMC block image, G-LMC block image and average block image, and a prediction error signal is produced from the difference between this selected block image and the image within a block of the original image. Therefore, the coding characteristics can be improved.