一种SAR图象的多方向多尺度融合边缘检测方法  

A Multidirection and Multiresolution Edge Detection Method for SAR Images

作　　者：李应岐[1] 田军 

机构地区：[1]西北工业大学电子工程系,陕西西安710072 [2]二炮驻骊山微电子公司军代室,陕西西安710075

出　　处：《微电子学与计算机》2005年第3期246-248,共3页Microelectronics & Computer

基　　金：国家自然科学基金(60072010;60272022)资助

摘　　要：本文从SAR图象的斑点噪声的特性分析出发,对以局部均值比ROA算子的统计特性进行了分析,给出了检测阈值和检测率的关系,提出了解决图象边缘过检测和欠检测的多方向多尺度融合检测算子,并将融合检测算子与其它Canny检测算子的检测结果进行了比较。The presence of a strong noise known as speckle complicates edge detection in Synthetic Aperture Radar(SAR) images and makes usual edge detection such as local maximum of the gradient magnitude inefficient. We here propose a new edge detector for SAR image based on the normalized ratio of local averages computed over opposite side of the central pixel in different directions and on different resolution levels infinite size windows. The false edges and the edge discontinuities can be suppressed. At last , the improvement compared to edge detector operating on SAR image is shown on.

关 键 词：SAR图象 边缘检测 融合 多方向 多尺度 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]