中文摘要：

为了在抑制相干斑噪声的同时更好地保持SAR图像中的点目标和边缘目标，在经典核回归方法的基础上，提出基于核回归的SAR图像自适应相干斑抑制方法。通过分析SAR图像的幅度分布特性，在构建模型时，以图像的幅度值为判别条件，使核函数在幅值较小的背景区域具有较大的光滑作用以抑制噪声，而在幅值较大的目标区域光滑作用较小以保护目标特征。同时考虑对边缘结构的保护作用，散射矩阵进一步修正自适应核回归方法的协方差矩阵，建立SAR图像自适应相干斑抑制方法。试验结果表明，该算法通过将幅度值和散射矩阵引入核函数，更好地抑制了噪声，同时也保护了图像中的点目标和边缘。

英文摘要：

An adaptive speckle reduction method based kernet regression for SAR image is presented in this paper, which can reduce speckle noise while preserving scatter targets and edges. Firstly, the image magnitude is introduced into the kernel function considering the characteristics on magnitude distribution of SAR image. Then the kernel function provides heavy smooth to reduce the speckle for background region, light smooth for targets region. Secondly, in order to preserve the edges effectively, the steering kernel is modified based on scatter matrix, And finally the algorithm for speckle reduction method based kernel regression for SAR image is summarized. The experiment results show that the proposed method can reduce speckle noise while preserving targets and edges by introducing magnitude information and scatter matrix into kernel function.