Region Segmentation of SAR Images Based on the Parametric Minimum Cut Method
Zhou Qiang-feng, Tian Zheng, Bingtao Liu · Gongcheng shuxue xuebao · 2010
The minimum cut criterion favors cutting small sets of isolated nodes in the graph,this paper propose a parametric minimum cut method to avoid it.Image segmentation based on graph method is not efficient when the number of the nodes is huge.This paper presents a novel graph-based SAR image region segmentation method with split-merge algorithm.The proposed method has two advantages,one is containing global information by using parametric minimum cut criterion to merge regions and the other one is greatly reducing nodes of the graph by using regions as nodes to create the graph.When splitting a SAR image into many regions according to its gray value,this paper gives a new homogeneity measurement of region to restrain speckle.The experiment results show the effectiveness of the approach.