Abstract: In the field of computer vision, the research on human motion analysis is attracting many researchers’ attention with its wide application prospect. The key technical issue is the human motion tracking in the image sequence. Because of the complexity of human motion, the existing research methods impose limitation on human. A new approach to track the human motion is brought forward in this paper. Instead of aiming at a given human motion mode, it can track large motion from frame to frame in a complex, variational background. The main scheme is to track every part of human body from top to bottom on the basis of human model. The image\|difference\|based motion estimation and the divided region histogram similarity algorithm are used. The final result is given at the end of the paper.