一种新的基于网格编码和区域合并的SAR图像快速分割算法

发布时间：2019-05-09 00:02

【摘要】：该文建立一种新的基于八邻域网格编码的SAR图像分割模型,并用区域合并技术实现了模型的快速求解。利用多方向比例边缘检测算子提取SAR图像的比例边缘强度映射(RESM),提出一种新的阈值处理方法抑制RESM均质区域内部的极小值,进而减少了对阈值处理后的RESM进行分水岭变换获得的初始分割的区域个数。递归地合并相邻区域来求取分割模型的次优解。利用区域邻接图(RAG)及其最近邻图(NNG)特性来加速区域合并过程。引入精确度(P)和召回率(R)来评价分割算法的边缘定位精度。与常用方法相比,该文方法具有高的边缘定位精度和低的时间复杂度。
[Abstract]:In this paper, a new SAR image segmentation model based on eight-neighborhood trellis coding is established, and the fast solution of the model is realized by region merging technique. The proportional edge intensity mapping (RESM),) of SAR images is extracted by multidirectional proportional edge detection operator. A new threshold processing method is proposed to suppress the minimum values in the RESM homogeneous region. Furthermore, the number of initial segmented regions obtained by watershed transformation of RESM after threshold processing is reduced. The sub-solution of the segmentation model is obtained by recursively merging the adjacent regions. The region adjacency graph (RAG) and its nearest neighbor graph (NNG) are used to accelerate the process of region merging. Precision (P) and recall (R) are introduced to evaluate the edge location accuracy of the segmentation algorithm. Compared with the common methods, this method has high edge location accuracy and low time complexity.
【作者单位】：西安电子科技大学雷达信号处理国家重点实验室;
【基金】：国家自然科学基金(61271295)资助课题
【分类号】：TN957.52

【参考文献】