A Tree-Like Multiphase Level Set Framework for 2-D and 3-D Medical Image Segmentation

The Chan-Vese model (C-V model) using one level set function is a two-phase active contour model and has limitation for medical image segmentation. In this paper, a tree-like multiphase level set framework based on the C-V model is proposed which applies the C-V model for multiple times to obtain multiple phases (n-1 times for n phases, n>1) and each operates as follows: First, a medical image is partitioned into an object and a background by the C-V model; Second, the pixels in the background is replaced by the mean of the object to create a new image (Hereafter, we name this step as the technique of painting background); Third, the new image is partitioned by the C-V model for the second time to obtain a new object. The proposed framework ensures that the new object is only inside the previous one. Experimental results for two dimensional (2-D) and three dimensional (3-D) medical images show that the framework is able to detect the objects with weak boundaries.