基于极化合成孔径雷达图像的斑点滤波方法

Speckle filtering method based on polarimetric SAR image

斑点噪声是由合成孔径雷达(SAR)的相干成像原理造成的固有缺陷,为了更好地进行SAR图像地形分类、目标检测等后续处理,提出一种用于极化SAR图像的滤波方法。该方法利用新的优化函数和约束条件,在保持图像功率等于L ee滤波功率的前提下,通过最小化信号子空间向量与原始向量的欧式距离,达到降斑的目的。实验结果表明:利用旧金山地区的真实极化SAR数据,使用该方法滤波的图像结果与原有的子空间滤波器相比,更接近原始图像的均值,并且通过滤波提高了不同类别的目标在特征空间的区分度,从而有利于不同类型地物的分类。

Speckles are inherent defects due to coherent imaging principles in synthetic aperture radar （SAR） systems. A polarimetric SAR image filtering method was developed using a new objective function and constraint conditions to get better results for terrain classification, target detection, and other applications. When the span of a filtered pixel equals to the span of a pixel filtered by Lee＇s method, the Euclidean distance between an unfiltered parameter vector and a signal space vector is minimized so that speckles can be reduced. Test results using polarimetric SAR data of San Francisco show that images filtered by this method have mean values closer to the unfiltered images than with a subspace based filter. After the filtering process, targets with different scattering characteristics are more separable in the characteristic space; therefore, this filter improves terrain classification.