Image Denoising with Rectified Linear Units.

Deep neural networks have shown their power in the image denoising problem by learning similar patterns in natural images. However, the traditional sigmoid function has shown its limitations. In this paper, we adopt the rectified linear (ReL) function instead of the sigmoid function as the activation function of hidden layers to further enhance the ability of neural network on solving image denoising problem. Our experiment shows that by better capturing patterns in natural images, our model can achieve better performance and less time consumption than those using sigmoid units. A large number of experiments show that our approach can achieve the state-of-the-art performance.