A Novel Supervised C-V Segmentation
Hai Zhang, Yi Zhen · 2011
Region-based active contours have attracted much attention in image segmentation, where Chan-Vese(C-V) model is a widespread algorithm. Many methods have been combined with C-V model and the common combination form is to use the results segmented by other methods as the input of C-V model. Unlike the published hybrid methods, a novel segmentation algorithm based on Otsu method and C-V model is presented. A new penalty function is defined to measure the difference between the level set function and the results segmented by Otsu method, which is added into the energy function as constraints to improve accuracy and speed of image segmentation. The comparison results illustrate that compared with C-V model, our algorithm is more insensitive to the initial zero level set and that our algorithm outperforms Otsu method and C-V model.