基于局部混合滤波的SAR图像边缘检测

该文结合基于非下采样方向滤波-双树复小波变换(NonSubsampled Direction Filter Bank-Dual-Tree Complex Wavelet Transform, NSDFB-DTCWT)的局部混合滤波算法和Dempster-Shafet (DS)证据理论提出一种基于局部混合滤波的SAR图像边缘检测算法。该算法首先对SAR图像进行局部混合滤波，然后对不同尺度滤波图像使用指数加权均值比(Ratio Of Exponentially Weighted Averages, ROEWA)算子检测边缘的强度，再使用Canny算子检测边缘的方向，从而得到SAR图像各尺度上的边缘，最后使用DS证据理论融合各尺度的边缘形成原始SAR图像的边缘。实验结果表明：该文所提出的算法具有很好的边缘检测效果，检测到的SAR图像的边缘定位准确和完整，且伪边缘较少。

SAR Image Edge Detection Based on Local Hybrid Filter

A SAR image detection method is proposed based on local hybrid filter. This method combines the advantages of both local hybrid filter based NonSubsampled Direction Filter Bank-Dual-Tree Complex Wavelet Transform (NSDFB-DTCWT) and Dempster-Shafet (DS) evidence theory. First, local hybrid filter is applied to SAR image. Then edge strength is got by using Ratio Of Exponentially Weighted Averages (ROEWA) operator and edge direction is got by using Canny operator at every scale from NSDFB-DTCWT. Finally, DS theory is used to fuse the edge in all scales to get the whole edge of the original SAR image. The experimental results demonstrate its effectiveness and superiority in terms of edge positioning accuracy and integrity, and it also has a fewer false edge points.