The block motion based on block matching is widely used for displacement estimation in motion-compensated hybrid coding. However, For low rate coding, the motion data is a big burden to coding efficiency.
In this thesis, the advantages of an object motion based on block motions is investigated for typical video-phone image of head and shoulder type. For each block of image, either block motion or object motion is selectively chosen by comparing displaced block differences. It is found that the motion data is significantly reduced but the overall coding efficiency is not improved much.