Bio-Computing Research Center, Harbin Institute of Technology, Shenzhen, Shenzhen, 518055, Guangdong, China.
Bio-Computing Research Center, Harbin Institute of Technology, Shenzhen, Shenzhen, 518055, Guangdong, China; Peng Cheng Laboratory, Shenzhen, 518055, Guangdong, China.
Neural Netw. 2020 Jan;121:461-473. doi: 10.1016/j.neunet.2019.08.022. Epub 2019 Sep 5.
Deep convolutional neural networks (CNNs) have attracted great attention in the field of image denoising. However, there are two drawbacks: (1) it is very difficult to train a deeper CNN for denoising tasks, and (2) most of deeper CNNs suffer from performance saturation. In this paper, we report the design of a novel network called a batch-renormalization denoising network (BRDNet). Specifically, we combine two networks to increase the width of the network, and thus obtain more features. Because batch renormalization is fused into BRDNet, we can address the internal covariate shift and small mini-batch problems. Residual learning is also adopted in a holistic way to facilitate the network training. Dilated convolutions are exploited to extract more information for denoising tasks. Extensive experimental results show that BRDNet outperforms state-of-the-art image-denoising methods. The code of BRDNet is accessible at http://www.yongxu.org/lunwen.html.
深度卷积神经网络（CNN）在图像去噪领域引起了广泛关注。然而，存在两个缺点：（1）对于去噪任务，很难训练更深的 CNN；（2）大多数更深的 CNN 存在性能饱和的问题。在本文中，我们报告了一种名为批量归一化去噪网络（BRDNet）的新型网络的设计。具体来说，我们结合了两个网络来增加网络的宽度，从而获得更多的特征。由于批量归一化被融合到 BRDNet 中，我们可以解决内部协变量偏移和小批量问题。还采用了整体的残差学习来促进网络训练。空洞卷积用于提取更多的信息用于去噪任务。广泛的实验结果表明，BRDNet 优于最先进的图像去噪方法。BRDNet 的代码可在 http://www.yongxu.org/lunwen.html 上获得。
