基于改进Chan-Vese模型的图像分割

Image segmentation based on improved Chan-Vese model

目前基于水平集的图像分割方法很难给出基于全局极值的算法终止条件,而大多采用事先设定迭代次数的方法.本文提出了一种改进的Chan-Vese模型,通过添加水平集函数约束项,使得新模型抑制了水平集函数的取值范围,最终收敛至全局极值,并以此作为算法终止条件,无需事先设定迭代次数.实验结果表明,新模型在其终止条件下,分割结果正确,与传统Chan-Vese模型相比,新模型的收敛速度快3～6倍,且通用性更强.

Most methods based on level set do not have the stop criterion based on the global minimum,instead they employ the iteration number as stop criterion.This paper proposes an improved Chan-Vese model,by adding a constrained term,making the functional converge to global minimum,and presents the stop criterion based on global minimum,without setting iteration number in advance.Experimental results show that the proposed model can segment images correctly under the new stop criterion.Compared with the Chan-Vese model,it has a faster convergence,3～6 times faster than Chan-Vese model,and is of good general use.