基于邻域均方连续差分的SAR图像边缘提取算法

摘要

该文提出了一种基于邻域均值连续差分平方和(均方连续差分)的SAR图像边缘提取算法.首先,将像素邻域滑动窗口分成多个互不重叠的子区域,采用邻域均值差分平方和作为边缘强度的衡量因子,从理论上证明了该算子可以消除区域亮度对边缘强度估计的影响,具有恒定虚警率的特性,同时较好地估计边缘方向.然后,根据边缘走向对边缘强度图像进行边缘细化,消除真实边缘附近的虚假边缘,并提出一种基于平均强度变化率的自适应双阈值连接方法提取SAR图像中边缘.仿真和实测SAR图像的实验结果表明,该文提出的算子在SAR图像的边缘检测中表现出较好的性能,具有较高检测率和边缘定位精度,边缘线段的连续性保持也较好.%An approach based on the square successive difference of neighborhood averages is proposed to extract edges in SAR images. The proposed operator partitions the neighborhood window of a pixel into many non-overlapping regions, and applies the mean square successive difference as an indicator of the edge strength. It is proved regardless of the mean intensity and is an edge detector with Constant False Alarms Rate (CFAR) theoretically. It also performs well in the estimation of edge orientations. The edge thinning algorithm and an adaptive double-thresholds processing are carried out to reduce false alarms nearby the true edges and to connect edges. The experiments on simulated and real SAR images indicate that the proposed operator achieves well performance in the edge detection and localization, and the extracted edge segments are long and continuous.