In this paper a novel algorithm is proposed for automatic segmentation of moving objects. Quick and precise abstraction of the moving objects enables a high level understanding and analysis of video contents and is very important in context based video coding (MPEG 4). Different from traditional approaches in which only luminace information is used
color information is also used in this algorithm as a reference in preprocess and the preprocess result is good. Traditional canny algorithm is improved here and is added a new ability to process two temporal correlative frames by which we get objects' refined but uncontinuous edges. An approach is proposed to get auxiliary edges by which refined edges are mended to be continuous edges of moving objects. Experimental results show that it's a fast and effective algorithm.