A Multi-phase Level Set Method for Image Segmentation Based on the Mumford-Shah Model

Image segmentation and contour detection based on the active contour is treated as the most popular hotspots. In this paper, a detailed analysis of Mumford-Shah Model is given with emphasis on its contour mechanism, and a simplified multi-phase piecewise smooth model is proposed, then a level set evolution without re-initialization is produced during the course of iteration. This new algorithm, whose initial curve is embedded in the level set function of higher dimension, can detect and segment correctly multiple objects of distinct intensities and handle the complex changes of topology. Finally, some experiments are conducted on the synthetic image and medicinal image, and satisfactory results show that this model is not sensitive to the initial location of the cure and is strongly robust to a noise.