Abstract: Based on that the active contour CV model lacks of the edge information, the CV model applied in color image appears some difficulties in the same and similar region. It extends the active contour model theory through the extended multi-stage active contour model to carry out the cluster, and integrates with the K means approach to determine the class number of the interactive division region automatically, meanwhile, obtains the density value between the inside and outside of evolution region, and joins the Geodesic Active Contour（GAC） that it is enable to capture the better ability of edge. Regarding to the optimization problem of image division, it can transform the optimization of energy function corresponding to the maximum-flow/minimum-cut problem. Through the experiment comparison in color image, it verifies the characteristic about effectiveness, high accuracy，along with litter over-weight burden computing of the approach proposed.

Key words: active contour, max flow-min cut, image segmentation

摘要： 基于活动轮廓的CV模型缺乏边缘的信息，使得CV模型在具有相同或相似区域的彩色图像分割应用时存在困难。对活动轮廓的模型进行扩展，提出了多段活动轮廓模型，应用K均值对交互区域进行聚类，确定分割区域中心的个数，同时得到多段活动轮廓模型进化边缘内外区域的密度值，加入测地线(GAC),使其具有更好的边缘捕获能力，对于图像分割的最优化问题，转化为求能量函数对应的图切分最大流最小切分问题。通过自然图像的实验，验证了提出的方法具有高效性、准确性、耗时少等特点。

关键词: 活动轮廓, 最大流最小切分, 图像分割