Variational Image Matting with Local and Nonlocal Operators

Image matting is a key procedure in image and video editing applications, ill-posed characteristic makes matting problems hard to solve. In this paper, we propose a novel variational image matting framework, and in this framework we apply both nonlocal operator and local operator to model the matting process. By using nonlocal operator, our method can work well with texture images. To get a stable numeric solution, we use the recent advance of convex optimization method in convex analysis domain. We test our method with a few natural image data sets, the results show effectiveness of the proposed framework.