基于CNN的图像滤境操作之去噪声
图像去噪声是图像增强、图像重建的重要内容之一，在现实中，很多图像难免存在噪声的问题，从而影响了图像的判别。当然图像去噪声在现有的图像处理软件中也比较成熟。
找到了一篇基于CNN的图像去噪声的论文，FFDNet: Toward a Fast and Flexible Solution for CNN based Image Denoising，对其结果进行复现，效果挺不错的。特记录如下：
在论文中主要记录了文章的三大贡献：
（1）A fast and flexible denoising network, namely FFDNet, is proposed for discriminative image denoising. By taking a tunable noise level map as input, a single FFDNet is able to deal with noise on different levels, as well as spatially variant noise.
• An orthogonal regularization method is proposed to improve the robustness of FFDNet to the mismatching
between input and ground truth noise levels.
• FFDNet exhibits perceptually appealing results on both synthetic noisy images corrupted by AWGN and realworld
noisy images, demonstrating its potentials on practical image denoising.
论文的架构主要如下：
其中CNN层架构在论文中描述如下：
从表述中可知，其框架结果也非常简单，其利用matconvnet的代码主要如下：
net = vl_simplenn_tidy(net);
res = vl_simplenn(net,input,[],[],'conserveMemory',true,'mode','test'); % 调用matconvnet 默认的函数
实验效果：（附图如下，可以看到，图像变得清晰）