基于Gomory-Hu算法有效实现的图像区域分割

提出一种基于图的图像区域分割方法。算法首先对原图像利用区域生长技术产生初始分割；其次以初始分割区域作为顶点构造赋权无向图；最后以Minimum Cut为准则，利用改进的Gomory-Hu算法得到图像的最终分割。该方法既减少了构造图的顶点又利用了全局信息来对区域分割。实验结果表明了该算法的有效性。

Effective implementation of image region segmentation based on Gomory-Hu algorithm

A graph-based image segmentation method was presented. Firstly, use region growing technique to find initial over-segmentation. Secondly, use these regions as nodes to create the graph. Finally, use minimum cut method to merge these regions. The proposed method has two advantages: one is using minimum cut criterion to merge regions, which can contain global information; the other is using regions as nodes to create the graph, which can greatly reduce nodes of the graph. The experimental results show the effectiveness of the approach.