一种基于SVR的SAR图像自适应滤波算法

针对合成孔径雷达（synthetic aperture radar, SAR）图像相干斑噪声抑制问题，提出了一种基于支持向量回归（support vector regression, SVR）分析的空间域自适应滤波方法。将SAR图像看做连续二维函数，利用SVR方法对其进行逼近。基于图像的逼近结果描述像素关联性，并基于关联性破坏程度对噪声进行类型分析，对不同类型的噪声采取确定性的抑制算法。为了保证精度，选择小波核函数构建支持向量回归机。实验结果表明了该方法的有效性和对经典方法的改进。

SAR speckle suppression algorithm based on SVR

To suppress the speckle in synthetic aperture radar (SAR) images, a novel adaptive algorithm based on SVR is proposed. The SAR image is regarded as a 2-D continuous function and is approximated by support vector regression (SVR). The result of regression is used to describe the relationship between pixels and the one in its neighborhood. Based on the relationship, the noise is classified to different kinds and suppressed. In the procedure of regression, the wavelet kernel function is used to improve the accuracy. The results of experiments show that this method is effective.