A novel approach for video object extraction fusing temporal and gradient information is proposed. Experiments demonstrate that our method can extract the video object effectively from complex background

and can resolve the vacancy phenomenon when the foreground and background are of the same color using background eliminate method. Firstly

in the temporal domain we introduced background eliminate and adjacent frame difference method to generate elementary video object

and then we used erosion and dilation morphologic method to the video object. Secondly

in the gradient domain we detected video object’s edge using Sobel operator. Associating with the video object generated in the temporal domain

we obtained the exact video object contour’s edge. Finally

we applied heuristic search method to link video object contour’s edge points

sequentially extracted video object.