基于Chan-Vese模型的树形结构多相水平集图像分割算法

由于Chan-Vese(C-V)模型通过单个水平集的符号将待分割图像划分为目标和背景两个部分,所以当图像的多个目标的轮廓成多连接时,C-V模型将无法表示.为了解决C-V模型在表示目标轮廓上的局限,提出了基于C-V模型的树形结构多相水平集算法.关键策略是通过改变图像背景,使得水平集在新图像上重新收敛;核心技术是依据同时明度对比提出的背景填充技术;算法流程采用多水平集串行收敛方式实现多相分割(n-1次收敛可以实现n相分割,n＞1).实验结果表明,本算法可以表示复杂的区域连接情况(n相分割最多可以表示n连接情况),能够实现多目标分割(n相分割可以实现n-1个目标分割),特别适合于目标中含有子目标的图像.

A Tree-Like Multiphase Level Set Algorithm for Image Segmentation Based on the Chan-Vese Model

The Chan and Vese(C-V) model using one level set function can only represent one object and one background by the sign of the function and can not express the multiple junctions of multiple objects.To deal with the problem,a tree-like multiphase level set algorithm for segmentation based on the C-V model is proposed whose basic idea is changing the background in the image so that the level set function will detect new object in the observed image.A key technique,called the technique of painting background,is proposed following the theory of simultaneous brightness contrast.Moreover,a hieranchical procedure of the proposed algorithm using multiple level sets is developed for multiphase segmentation(n-1 level sets for n phases,n>1).Experimental results show that the proposed algorithm can represent multiple junctions of regions(n-phase segmentation for at most representing n junctions) and detect multiple objects(n-phase segmentation for n-1 objects).Also,the algorithm is especially effective for images which have sub-objects in the object region.