基于M-S模型的参数自适应图分割方法

基于图分割的分段常数水平集方法Mumford-Shah模型能够很好的处理多相图像，但由于Mumford-Shah模型本身固有的原因，该方法对于长度参数也很敏感。为了解决这个难题，本文给出了一个自动估计包括长度参数的模型参数的自适应方法，该方法根据图像的噪声方差大小来确定参数。该自适应方法与基于图分割方法相结合，能够解决模型对参数敏感的难题。最后，通过计算机生成的图像和实际图像的分割结果证明了该方法是有效的。

Segmentation of Image ,Based on Graph Cuts＇ M-S Model by Using Parameter Estimation

The Mumford - Shah based on graph cuts＇ approach by using piecewise constant level set method is a very effective model for multi-phase images, sensitive to the parameter of length because of reasons inherent in the Mumford-Shah model itself. To solve this problem, a adaptive method is given, which can automatically estimate model parameters including the length of the parameter defined as an increasing function of the variance of the noise. The problem of model parameter sensitivity can be solved by the adaptive method and the combination of graph-based segmentation method. Finally, segmentation results of computer-generated images and real image show that the method is effective.