Local Energy-Driven Active Contour Model with a Global Minimum

摘要

The Mumford-Shah model is a powerful and robust segmentation technique. However ,the numerical method of solving the Mumford-Shah model is difficult to implement, and often gets into a local minimum. In this paper,we propose a novel active contour model for image segmentation based on Chan-Vese model, in which the curve driven by local energy within a window around the curve,can converge to a global stationary minimum. The novel active contour model utilizes fully the region and boundary information in the whole image domain. In the new energy functional of our model, the modified data fitting term can guarantee a global stationary solution, and the alignment term can help to improve the speed of cure evolution. The experiment results on synthetic and medical MRI images show that, our novel active contour model obtain a stable segmentation result and the final result does not depend on the iteration number of algorithm.