改进模糊马尔可夫随机场的SAR图像分割

SAR Image Segmentation Based on Improved Fuzzy Markov Random Field

模糊马尔可夫随机场利用模糊隶属度函数解决了马尔可夫随机场过于依赖灰度值的弊端,然而在模糊马尔可夫随机场的分割中采用的是硬性基团和边缘判断准则,模糊分类使得这些判断条件更加难以满足,从而图像分割中容易产生边缘分割错误的现象。改进模糊马尔可夫随机场是在分段模糊马尔可夫随机场中引入模糊意义的后验概率公式及基团和边缘类型"模糊相似性"概念,通过对不同基团和边缘类型的模糊相似性描述与判断,使得图像分割对于边缘的判断和噪声的抑制具有更好的效果。仿真实验表明改进模糊马尔可夫随机场对于边缘特别模糊的合成孔径雷达图像(SAR)具有较好的分割效果。

Fuzzy Markov Random Field （FMRF） decreased the dependency on the gray scale in the MRF image segmentation by using the belief function. However the strict criterions of cliques and borders in FMRF were hardly accorded in many neighborhoods, the segmentation results had many mistakes in the borders of regions and the relative noise. This paper presents an improved FMRF segmentation method based on the pairwise FMRF and introduced the fuzzy similarity of cliques and borders which could distinguish the energy contributions by the different cliques or borders that were stained by noise. The improved FMRF was more sensitive than others FMRF and had the strong ability to restrain the noise and recognize the borders. Finally the experiments demonstrate that the improved FMRF algorithm is efficient to segment SAR images.