An automatic video object segmentation scheme

In this paper, a new scheme is presented for performing automatic video object segmentation. First of all, a new adaptive and reliable threshold estimation method has been proposed for the change detection process. Secondly, the initial edge model of the video object is obtained on the basis of the adaptive change detection and Canny edge detection. Thirdly, the edge model is tracked and updated in every frame to accommodate rotation and changes in shape of the tracked object and then the outer contour of the video object is determined based on the mathematical morphological operation. Lastly, an improved active contour which uses the gradient vector as the external force is applied to guide the initial outer contour moving to the actual video object contour. Experiment results show that the proposed algorithm is robust to the entire motion and local deformation of object, adaptable to the complex background and can get the closed and accurate video object contour.