An Incremental Chan-Vese Model for Fast Medical Image Segmentation

Chan-Vese model is a powerful and flexible method for many fields of computer vision such as image segmentation. However, segmentation with this method involves a lot of iterations on the whole image region, which leads to a low calculating efficiency, especially for medical image segmentation. In this paper, a new incremental Chan-Vese model for fast image segmentation is presented. We define two concepts: signed member function and signed area, and then establish a progressive iterative formula for Chan-Vese model to calculate the average value of a whole image region during every iteration. Based on this incremental model, some fast evolution algorithms such as the narrowband method can be adopted to improve the efficiency of segmentation. This model is applied to segmentation of 3D medical images, and experiments confirm its high performance, which makes classical Chan-Vese model more practical.