基于MSP-ROA边缘检测的SAR图像分割

SAR Image Segmentation Based on MSP-ROA Edge Detection

该文以卡通模型为SAR图像分割的理论依据,提出了一种基于MSP-ROA边缘检测的SAR图像分割算法,算法中对边缘检测的结果做种子生长、标注和区域填充,并根据相似性准则对填充的初始分割结果进行相邻区域的合并处理,最终得到同质性和连通性都较好的SAR图像分割结果;文中给出了SAR图像分割处理的性能评估指标,并用实测的SAR图像数据验证了上述算法,最后给出了与其它分割方法比较的结果。

This paper firstly introduces the cartoon model which SAR image segmentation is based on. A new method of SAR image segmentation which based on MSP-ROA edge detection is proposed. After edge detected, the algorithms including seed growing, region signing and region filling are used to have a good image segmentation in which the segments have good connectivity and homogeneity. Objective methods for evaluating of the segmentation algorithm＇s performance are discussed, and the real SAR image data are applied to validate the above algorithm. Finally, the method in this paper is compared with the other segmentation algorithm.