Synergistic integration of graph-cut and cloud model strategies for image segmentation

This paper proposes a new graph cut image partitioning method that calculates image data using cloud model for constructing the objective functions (GC-CM). In the objective function, it contains a boundary preserving smooth term and a data item which evaluates the deviation of each pixel that belongs to different regions. The core method models the foreground object and background of the images as cloud models by the back cloud generator. The data item is calculated with the X-condition cloud generator. We use the membership degree between each pixel to calculate the similarity of the neighbor pixel established as the smooth term. The energy minimization is completed with the minimum cut theory and the graph cut iterations. In contrast to segmentation results with discontinuous edges using conventional graph cut method, this method has better generality and accuracy. Experiments on different data sets including natural images from Berkeley database, synthetic data, and medical images suggest that the proposed method based on cloud model and graph cuts outperforms other state-of-the-art approaches.