Fast Multiregion Image Segmentation Using Statistical Active Contours

In this letter, we propose a novel statistical active contours method for using an arbitrary number of level set functions to segment an image into regions of the corresponding amount. First, a new updating of level set functions for every region is derived from the probabilistic models of image data. Second, a novel geometric prior that deduced from the level-set-based curve evolution is introduced to obtain the probabilistic label. Therefore, updating of the level set functions and estimation of the distribution parameter are run alternately in a fast manner. Finally, in order to further enhance the efficiency, we initialize the level set function by the statistical approach to draw near object boundary. We experimentally evaluate our proposed approach on complicated real-world images and demonstrate its good performance in practice.