利用区域划分的合成孔径雷达图像相干斑抑制算法

A Despeckling Algorithm for Synthetic Aperture Radar Images Using Region Subdivision

为在有效保护合成孔径雷达(SAR)图像边缘特征的同时进一步提高乘性相干斑噪声的抑制性能,提出了一种基于区域划分的SAR图像相干斑抑制(DABRS)算法.首先利用带方向的高斯-伽马平行窗通过比率运算生成边缘强度映射(ESM),并利用阈值化ESM对SAR图像进行区域划分,然后利用改进的变窗Kuan滤波与改进的Sigma滤波分别对SAR图像的均匀区域与边缘区域进行相干斑抑制.为进一步提高抑斑效果,两种改进抑斑算法均采用迭代滤波方式,且改进Kuan滤波的窗尺度随迭代次数的增加而增大.实验结果表明:与多种抑斑算法相比,DABRS算法在抑斑与边缘保护方面具有优势,等效视数提高超过30%,边缘保持指数提高8%以上,而且抑斑图像同质区还具有更平滑的视觉效果.

A new despeckling algorithm based on region subdivision（DABRS） is presented to effectively preserve edge characteristics and to improve the performance of reducing multiplicative speckle noise in synthetic aperture radar（SAR）.Gaussian-Gamma-shaped bi-windows with different orientations are used to obtain the edge strength map（ESM） by ratio operations.Then,the threshold-processing ESM is used to divide a SAR image into different regions.Two improved filters,the Kuan filter with variable-size windows and the Sigma filter,are used to reduce speckle in homogeneous and edge regions,respectively.Iteration filtering strategy is adopted by the two improved filters to smooth speckle better,and the window size of improved Kuan filter increases with iteration progressing.Experimental results and comparisons with most existing despeckling algorithms show that the DABRS has advantages in the speckle suppression and edge preservation,improves the equivalent number of looks by more than 30%,increases the edge keep index by more than 8%,and has more smooth visual effects in the homogeneous regions of despeckling images.