基于核独立成分分析的极化SAR图像相干斑抑制

Research on Polarimetric SAR Image Speckle Reduction Using Kernel Independent Component Analysis

为提高极化合成孔径雷达图像相干斑抑制的效果,提出基于核独立成分分析(kernel independent component analysis,KICA)的极化SAR图像相干斑抑制方法。该方法将三个通道的极化信息作为输入数据,经过KICA变换得到三个独立分量,取相干斑指数最小的分量作为滤波后的信息图像。由于将核函数引入到独立成分分析(independent component analysis,ICA)中,使在ICA中无法线性可分的信息在高维空间中达到线性可分。采用旧金山地区的AIRSAR数据与日本新潟地区的PISAR数据分别进行试验,并用相干斑指数和边缘保持系数从客观上进行评价。试验表明,与ICA算法相比,KICA算法具有更好的滤波效果和保持边缘信息的能力。

In order to improve the accuracy of polarimetric synthetic aperture radar image speckle reduction,a polarimetric SAR image speckle reduction method using kernel independent component analysis（KICA） is presented.This method uses the polarimetric information of three channels as its input data,obtains three independent components after KICA conversion,and takes the one with the smallest speckle index as the filtered results.Due to the introduction of kernel function,the information that can not be linearly separated using independent component analysis（ICA） algorithm achieves linearly separated in the kernel high-dimensional space.For the purpose of verifying the validity of the KICA method,the AIRSAR data of San Francisco and the PISAR data of Japan＇s Niigata were tested.The efficiency is objectively evaluated by the speckle reduction index and the edge preservation index.And the experiment results show that the image edges are retained better and the speckles are removed more effectively with the method of KICA algorithm compared with the ICA algorithm.