快速加权核范数最小化的SAR图像去噪算法

提出快速加权核范数最小化(fast weighted nuclear norm minimization,FWNNM)的合成孔径雷达(synthetic aperture radar,SAR)图像去噪算法。首先采用对数变换将SAR图像的乘性噪声变换为加性噪声,然后利用非局部相似性对变换后的图像进行块匹配,随后根据低秩模型框架,用随机奇异值分解替换加权核范数最小化(weighted nuclear norm minimization,WNNM)算法中的奇异值分解进行低秩矩阵逼近,再采用梯度直方图保存的方法对图像进行纹理增强,最终实现了对SAR图像快速去噪。在MSTAR数据库上的实验结果表明,与已有方法相比,所提方法在SAR图像去噪和边缘保持方面是有效的,并且比WNNM去噪速度快3倍。

SAR image denoising via fast weighted nuclear norm minimization

A synthetic aperture radar (SAR) image denoising method based on fast weighted nuclear norm minimization is proposed. Firstly, multiplicative speckle noise is converted to additive noise by logarithmic transformation. Secondly, the nonlocal similarity is used for image block matching. Next, according to the framework of the low rank model, random singular value decomposition is introduced to replace the singular value decomposition in the weighted nuclear norm minimization (WNNM) algorithm for approximating the low rank matrix. Then to enhance the texture of the image, the gradient histogram preservation method is used. Finally, fast denoising of SAR images is achieved. Experiments on the MSTAR database show that the proposed approach is effective in SAR image denosing and the edge preserving in comparison with some traditional algorithms. Moreover, it is three times faster than the WNNM method.