Enhanced CNN for image denoising

TL;DR

This paper introduces ECNDNet, a novel CNN architecture for image denoising that uses residual learning, batch normalization, and dilated convolutions to improve training efficiency and denoising performance.

Contribution

The paper proposes ECNDNet, a new CNN model that combines residual learning, batch normalization, and dilated convolutions to enhance image denoising.

Findings

ECNDNet outperforms existing denoising methods in experiments.

Residual learning and batch normalization improve training speed.

Dilated convolutions enlarge context information with reduced computation.

Abstract

Owing to flexible architectures of deep convolutional neural networks (CNNs), CNNs are successfully used for image denoising. However, they suffer from the following drawbacks: (i) deep network architecture is very difficult to train. (ii) Deeper networks face the challenge of performance saturation. In this study, the authors propose a novel method called enhanced convolutional neural denoising network (ECNDNet). Specifically, they use residual learning and batch normalisation techniques to address the problem of training difficulties and accelerate the convergence of the network. In addition, dilated convolutions are used in the proposed network to enlarge the context information and reduce the computational cost. Extensive experiments demonstrate that the ECNDNet outperforms the state-of-the-art methods for image denoising.