Kernel density feature based improved Chan-Vese Model for image segmentation

In this paper, an interactive image segmentation method is proposed base on the kernel density feature estimation. Compared with the traditional RGB value, it could be more accurate to model the color feature of pixel using corresponding kernel density estimation. To obtain the regional color feature, the mean of kernel densities of all pixels in this region is applied, and Bhattacharyya distance is used to measure the differences between two kernel densities. Consequently, an energy function is constructed according to the main idea of Chan-Vese Model, and it is optimized using the graph cuts technique. Experimental results demonstrate the advantages of our proposed method in terms of robustness and accuracy, especially for objects with thin elongated or concave parts.