区域化模糊聚类SAR图像分割

Regionalized Fuzzy Clustering SAR Image Segmentation

将图像域规则划分与模糊聚类方法结合,提出了一种区域化模糊聚类算法,并将该算法用于合成孔径雷达(Synthetic Aperture Radar,SAR)图像分割,以解决分割过程中像素模糊聚类难以处理SAR图像中存在的大量固有斑点噪声问题。首先,利用规则划分技术将图像域划分成大小相等的规则子块;假设每一子块内像素对聚类的隶属度相同,并以此为基础定义区域模糊聚类目标函数;通过迭代最小化上述目标函数实现SAR图像初步分割;最后,采用中值滤波方法进行后处理操作,以消除规则划分对不同类别之间边界的影响,实现SAR图像精准分割。为了验证提出算法的有效性,用模拟及真实SAR图像实现了算法测试;对算法分割结果进行定性与定量评价。结果表明算法的分割精度较高,可以有效降低SAR图像中斑点噪声对分割结果的影响。

In this paper,we combine the image domain rule classification with fuzzy clustering method,a region based fuzzy clustering algorithm is proposed,and the algorithm is applied to the segmentation of SAR images to solve the problem of large number of inherent speckle noise in SAR images which are difficult to be processed by fuzzy clustering in segmentation process. Firstly,the image domain is divided into regular sub blocks by rule partition and it is assumed that each sub block has the same degree of membership,based on this,the objective function of fuzzy clustering is defined; through the iteration,minimizing the above objective function to achieve SAR image segmentation; Finally,the median filter is used to carry out the post-processing,in order to eliminate the influence of rule division on boundary between different classes and to achieve the final segmentation of SAR images,to achieve accurate segmentation of SAR image. In order to verify the effectiveness of the proposed algorithm,the algorithm is tested by simulation and real SAR images; qualitative and quantitative test results show that the segmentation accuracy of the algorithm is high,It can effectively overcome the influence of speckle noise in SAR images on the segmentation results.