Abstract: In order to enhance the visual perception of image denoising, this paper proposes a method named two image denoising method combining Total Variation（TV） with Weighted kernel Norm Minimization（WNNM）. The noisy image is denoised with TV, then the noisy image and the based denoised one are expected to be made a differential operation. After that the result will be filtered with the adaptive Wiener filter. Having been filtered, the image will be overlain with the TV based denoising image, and be made the similar patches collected by using block matching. The final denoised image will be formed after the twice denoising with WNNM. Compared with the original WNNM, Block Matching 3-D （BM3D） and Foveated NL-Means（FNLM）, this method can make a better denoising effect on smooth areas; meanwhile, it also can reduce the spots and false fringe status which are caused by FNLM and BM3D under the high noise. The structure and texture information can be furthest similar as well.

Key words: weighted kernel norm, total variation, image residual, twice denoised

摘要： 为提升图像去噪后的视觉感受，提出一种加权核范数最小化（WNNM）结合全变分（TV）的二级图像降噪方法。首先对含噪图像进行TV基础去噪，其次用噪声图像与基础去噪结果图做差分运算，并对差分后的结果自适应维纳滤波，然后将滤波后图像与基础TV降噪图像叠加，利用块匹配做相似补丁收集，最后运用加权核范数最小化进行二次去噪，得到最终降噪图像。通过与原WNNM、三维块匹配去噪（BM3D）、漏斗自相似非局部去噪（FNLM）方法对比，该方法不仅对平滑区域有较优的降噪效果，同时处理了漏斗自相似非局部去噪与BM3D在高噪声情况下带来花斑与假条纹状况，并且使结构纹理信息最大化相似。

关键词: 加权核范数, 全变分, 图像残差, 二次去噪