A Markov Random Field Model-Based Approach to Natural Image Matting

This paper proposes a Markov Random Field (MRF) model-based approach tonatural image matting with complex scenes. After the trimap for mattingis given manually, the unknown region is roughly segmented into severaljoint sub-regions. In each sub-region, we partition the colors ofneighboring background or foreground pixels into several clusters inRGB color space and assign matting label to each unknown pixel. Allthe labels are modelled as an MRF and the matting problem is thenformulated as a maximum a posteriori (MAP) estimation problem.Simulated annealing is used to find the optimal MAP estimation. Thebetter results can be obtained under the same user-interactions whenimages are complex. Results of natural image matting experimentsperformed on complex images using this approach are shown andcompared in this paper.