A Multi-scale SAR Segmentation Based on Hierarchical Merging
SAR image segmentation is the base problem of SAR processing,and the premise of the scene interpretation.Be corrupted by speckle noise,SAR image segmentation is a recognized problem.The paper presents an approach using the method based on hierarchical region merging.According to the mechanism of speckle,the backscatter of SAR image obeys the negative exponent distribution in homogeneous region.We proposed the measure of the homogeneity based on negative exponent distribution and Mumford-Shah function.The Local mutual best fitting,which has lower algorithm complexity,will be used in our algorithm.Unlike the traditional hybrid segmentation,the method presented in our paper can afford the segmentation result in continuous scales from pixel level to the whole scene.The comparison of the method and Level Set validates that segmentation method can overcome the speckle noise and better in keep the structure of the parcel.
Xi Ye Meng Liu Hong Zhang Chao Wang
Center for Earth Observation and Digital Earth,CAS,Beijing 100094,China;Graduate University of the C Center for Earth Observation and Digital Earth,CAS,Beijing 100094,China
国际会议
Progress in Electromagnetics Research Symposium 2013(2013年电磁学研究新进展学术研讨会)
台北
英文
251-254
2013-03-01（万方平台首次上网日期，不代表论文的发表时间）