A Convex Approach for Local Statistics Based Region Segmentation

A convex active contour model based on local image statistics is proposed in this paper. By assuming that the intensity distribution of the image pixels in a window is described by a Gaussian distribution, our model is able to segment images with intensity inhomogeneity. Due to the convexity of the proposed model, we introduce a dual formulation to solve the minimization problem and obtain a much efficient method. Experiments show that the segmentation results of the proposed method are similar to that of the non-convex method based on local statistics, but our method is much more efficient.