Initialization Techniques for Chan-Vese Model with Thresholding

Abstract

Chan-Vese model is one of classical active contour models for segmentation based on level set methods. It is the region-based model but in some cases it is still sensitive to the location of initial contours. The image thresholding is a simple but effective tool to separate objects from the background. In this paper, we integrate these two techniques and propose a new method to improve the initialization for Chan-Vese Model. First analyze the distribution of image gray level histogram and find the optimum threshold values, then set the model’s initial contours with thresholds and construct energy functional, lastly iterate the functional formulations until convergence to the object boundary. The method is tested on the plaque images and gives considerable increase in performance.