Efficient Foreground Segmentation Using an Image Matting Technology

Efficient segmentation of foreground moving objects is an important procedure to achieve stable object tracking and recognition. In this paper, we have developed a novel algorithm for foreground segmentation that can extract moving objects from background accurately and efficiently. In our proposed algorithm, Gaussian Mixture Model is first used to model the static background regions. The boundary box of the foreground regions is determined via inter-frame change detection and SIFT feature analysis, and Grabcut algorithm(an image matting technology) is then used to obtain the optimal segmentation of foreground moving objects. Experiments on a set of video clips with huge diverse scenes have demonstrated the efficiency of our proposed method.