几种基于参数估计的图像分割比较

基于变分水平集方法建立了两相图像分割的通用模型，其区域模型基于图像的一般概率分布函数，通过变分方法得到水平集函数的演化方程和符合常值分布、Gauss分布及Rayleigh分布的参数估计。通过半隐式差分格式对演化方程离散求解，对含不同噪声分布图像采用不同的参数估计模型进行分割，得到了有意义的结论，得出三种模型适用范围，并分析各个模型驱动曲线演化所依据的图像蕴含信息。此外，通过在能量泛函中增加使水平集函数为符号距离函数的约束项，有效地避免了水平集函数需要重新初始化的问题。

Comparison for image segmentation based on different parameters estimate

A general model of two phase image segmentation based on variational level set method is proposed，and the region model is based on the image general probability distribution function.The evolution equation of level set function and the parameters’ estimate of piecewise constant values，Gaussian and Rayleigh distribution are obtained through the variational method.The evolution equation is discreted and solved through the semi-implicit finite difference scheme.Several useful examples are given for images segmentation，and some useful guidelines are summarized for application on the different distribution images；each model that drives curve evolving according to the information of images is analyzed.In addition，constraints of the level set function as signed distance function are enforced in the energy functional to avoid re-initialization.