一种新的非线性干涉SAR噪声抑制方法

现有的常用于干涉SAR相位图噪声抑制的非线性滤波法主要有模糊中值滤波和模数滤波，这两种方法都会引起比较大的图像边缘模糊，从而降低后续处理的精度。邻域平均法是一种非常经典的图像降噪方法，它具有较好的边缘保持特性。论文结合干涉SAR相位图的特性，改进了邻域平均法的邻域甲均值计算公式，有选择地利用邻域平均值代替当前像素点的值，提出了一种适用于干涉SAR相位图降噪的新的非线性滤波方法一条件邻域平均法，它具有更好的边缘保持特性。在理论分析和比较条件邻域平均法与模糊中值滤波和模数滤波方法的基础上，采用ERS1／2干涉SAR数据对论文提出的方法和所作的分析进行了验证。

A Novel Nonlinear Filter for Noise Suppressing of Interferometric SAR

The nonlinear filter is an important method for phase image noise reducing of interferometric SAR. The modified median filter and the modified mode filter are the main nonlinear filters to reduce the noise of interferometric phase image. They can suppress the noise well, but blur the image edge. The neighbour average is a kind of noise suppressing methods which keep the image edge well. Based on the neighbour average method, a novel nonlinear filter was presented to suppress the phase noise of interferometric SAR and it was referred as the modified neighbour average filter which can keep the image edge well. Combining the phase image characteristics of interferometric SAR, the new formula to calculate the average value was presented and some rules were applied to limit the filtering. The modified neighbour average filter was compared with the modified median filter and the modified mode filter. With the Residual point concept, one method was given to evaluate the phase noise suppressing methods. Finally, ERS1/2 interferometric data was processed with the three nonlinear filters and the results were employed to illustrate the correction of the analysis of this paper.