基于预筛选的改进的SAR图像PPB去斑

Improved probabilistic patch-based SAR image despeckling based on preselection

针对迭代式PPB滤波器时耗过长,而其非迭代式滤波器PPB(non-it)抑制了滤波后SAR图像的纹理和细节特征以及因没有对像素点进行有效的预筛选而影响图像视觉效果的矛盾,提出了一种基于预筛选的改进的SAR图像PPB去斑方法。该方法利用目标反射强的特征对其进行检测和保护,同时缩小滤波器在该处搜索窗口和比较窗口的尺寸;对于其他的像素点,通过像素块均值和方差的比较剔除搜索窗口中不相关的像素点,结合改进的sigma滤波器技术,利用sigma范围进一步排除异值点,让剩余的像素点参与中心点的滤波。实验结果表明,与原PPB(non-it)滤波器相比,改进的滤波器在图像纹理、细节保持和斑点抑制能力方面都有显著的提高,取得了较为满意的效果。

An improved probabilistic patchbased （PPB） SAR image despeckling based on preselection is proposed for the PPB （iterative） filter＇s disadvantage of timeconsuming and the PPB （nonit） filter＇s suppression of image details and texture and un satisfactory visual quality without effective preseleetion. Firstly, highreturn targets are detected and preserved, in which the sizes of search window and similarity window are reducedfor the rest of the pixels, the unrelated pixels in the search window are eliminated by comparing the means and the variances of patches, then incorporating the technique of improved sigma filter, the outliers are excluded further by the sigma range. Finally, only the rest of the pixels in the search window can attend the despeckling. Experimental results illustrate that the improved filter performs much better in details and texture preservation and suppression of speckle in comparison with the original PPB （nonit） filter and the results are satisfactory.