An adaptive video segmentation approach based on shape prior

As the basic and efficient segmentation framework, GraphCut plays an important part in video segmentation area. This paper proposes a adaptive video segmentation approach based on shape prior of the foreground. Shape information with Euclidean distance measure is added to GraphCut framework to compensate instability caused by single color information. And the shape model is adaptive with the size of foreground. The experiments show segmentation results with our method is significantly better than only using the color information.