Comparing Chan-Vese model and convex Chan-Vese model

摘要

Chan-Vese model, i.e., active contours without edges, is a successful variational model for segmentation. However, it is also a non-convex model, which means that quality of the segmentation result depends on the choice of initialization. An approximating model, in which the energy function is convex, ensures that there is a unique optimal segmentation and the model is then called convex Chan-Vese model. In this paper, we discuss Chan-Vese model and convex Chan-Vese model thoroughly and point out that Convex Chan-Vese model is outstanding. The conclusion is based on the analysis and comparison of experimental data.