Automatic video object segmentation using wavelet transform and improved active contour

In this paper, an automatic and accurate video object segmentation algorithm based on wavelet transform and improved active contour is proposed. Firstly, the wavelet transform is applied to two consecutive frames. Then, change detection method and Canny edge detection are applied in wavelet domain. After merging the different edge masks of four sub-bands with union operation, the robust difference edge map of the inter-frame can be obtained. Secondly, the current frame´s moving edge is detected and tracked by the difference edge map, current frame edge map, background edge map and previous frame´s moving edge. Lastly, morphological operations are applied to the current frame´s moving edge to get the outer contour of the moving object and an improved active contour which uses the gradient vector as the external force is used to guide the initial outer contour moving to the actual video object contour. Experiment results show that the proposed algorithm is effective in precise segmentation.