Fast and automatic segmentation of video object is a key technology in object based video coding such as MPEG 4. In this paper

a new algorithm for extracting moving object based on spatio temporal information is proposed. First

a binary motion image is achieved based on high order statistics detection method and motion information of multiple frames. High order statistics detection method can get rid of noise and disturbance effectively. Then

an improved watershed algorithm is proposed to segment motion region and its surrounding areas. Compared with traditional watershed algorithm the improved watershed algorithm can reduce the over segmentation phenomenon as well as the computational load. Finally

a moving object is extracted with projecting operation of spatial and temporal segmented results. In order to test the performance of the proposed algorithm

simulation test is performed with test sequences Sussie and Missa. And good experimental results are acquired which show that the proposed algorithm is efficient.