自适应堆栈滤波在极化SAR图像分类预处理中的应用

Application of Adaptive Stack Filter in Preprocessing of Polarization SAR Image Classification

堆栈滤波器属于非线性滤波器的一种,它在滤波后图像的边缘和细节保持方面有很好的效果。研究将自适应堆栈滤波器引入到对极化合成孔径雷达(pol-SAR)图像分类的预处理,即滤除极化SAR图像相干斑噪声。而递归过程在平均绝对误差(MAE)准则下对堆栈滤波器的优化也在缩短滤波程序执行时间和强化输出图像质量方面都有不错的效果。滤波后的图像将会首先进行相干数据分解(H/A/α分解),基于分解后的数据使用H-α/Wishart分类方法来进行极化SAR图像分类,分类效果将会与基于其他滤波预处理后的分类图像进行比较并得出最后结论。

As a special case of non - linear filters, stack filters have a good performance for filtering images with different types of noise while preserving edges and details. In this paper, a adaptive stack filter is designed for the preprocessing of classification of polarimetric Synthetic Aperture Radar （ pol - SAR） images, which are affected by speckle noise. The recursive implementation approach of optimal stack filters under MAE criterion is advanced in shorting the running time of the filtering procedure and improve the quality of the output. The results from the output of stack filter are extracted into polarized characteristics firstly by using the H/A/α decomposition method. And based on these characteristics, applying H- α/ Wishart methods will help us to classify the information in the SAR image. The effects of classification are evaluated and then are compared with the results of classifying the images with other methods of filtering.