Enhanced CNN for image denoising
Owing to the flexible architectures of deep convolutional neural networks (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....
Main Authors: | , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Wiley
2019-01-01
|
Series: | CAAI Transactions on Intelligence Technology |
Subjects: | |
Online Access: | https://digital-library.theiet.org/content/journals/10.1049/trit.2018.1054 |
id |
doaj-6e18a2f9aab344918a5ee0523cda5f96 |
---|---|
record_format |
Article |
spelling |
doaj-6e18a2f9aab344918a5ee0523cda5f962021-04-02T09:20:44ZengWileyCAAI Transactions on Intelligence Technology2468-23222019-01-0110.1049/trit.2018.1054TRIT.2018.1054Enhanced CNN for image denoisingChunwei Tian0Yong Xu1Lunke Fei2Junqian Wang3Jie Wen4Nan Luo5Bio-Computing Research Center, Harbin Institute of TechnologyBio-Computing Research Center, Harbin Institute of TechnologySchool of Computer Science and Technology, Guangdong University of TechnologyBio-Computing Research Center, Harbin Institute of TechnologyBio-Computing Research Center, Harbin Institute of TechnologyInstitute of Automation Heilongjiang Academy of SciencesOwing to the flexible architectures of deep convolutional neural networks (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.https://digital-library.theiet.org/content/journals/10.1049/trit.2018.1054image restorationimage representationlearning (artificial intelligence)image denoisingneural netsconvolutionimage restoration CNNimage denoisingenhanced CNNflexible architecturesdeep convolutional neural networksdeep network architectureDeeper networksperformance saturationconvolutional neural denoising networkresidual learningbatch normalisation techniquestraining difficultiesdilated convolutionsauthors |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Chunwei Tian Yong Xu Lunke Fei Junqian Wang Jie Wen Nan Luo |
spellingShingle |
Chunwei Tian Yong Xu Lunke Fei Junqian Wang Jie Wen Nan Luo Enhanced CNN for image denoising CAAI Transactions on Intelligence Technology image restoration image representation learning (artificial intelligence) image denoising neural nets convolution image restoration CNN image denoising enhanced CNN flexible architectures deep convolutional neural networks deep network architecture Deeper networks performance saturation convolutional neural denoising network residual learning batch normalisation techniques training difficulties dilated convolutions authors |
author_facet |
Chunwei Tian Yong Xu Lunke Fei Junqian Wang Jie Wen Nan Luo |
author_sort |
Chunwei Tian |
title |
Enhanced CNN for image denoising |
title_short |
Enhanced CNN for image denoising |
title_full |
Enhanced CNN for image denoising |
title_fullStr |
Enhanced CNN for image denoising |
title_full_unstemmed |
Enhanced CNN for image denoising |
title_sort |
enhanced cnn for image denoising |
publisher |
Wiley |
series |
CAAI Transactions on Intelligence Technology |
issn |
2468-2322 |
publishDate |
2019-01-01 |
description |
Owing to the flexible architectures of deep convolutional neural networks (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. |
topic |
image restoration image representation learning (artificial intelligence) image denoising neural nets convolution image restoration CNN image denoising enhanced CNN flexible architectures deep convolutional neural networks deep network architecture Deeper networks performance saturation convolutional neural denoising network residual learning batch normalisation techniques training difficulties dilated convolutions authors |
url |
https://digital-library.theiet.org/content/journals/10.1049/trit.2018.1054 |
work_keys_str_mv |
AT chunweitian enhancedcnnforimagedenoising AT yongxu enhancedcnnforimagedenoising AT lunkefei enhancedcnnforimagedenoising AT junqianwang enhancedcnnforimagedenoising AT jiewen enhancedcnnforimagedenoising AT nanluo enhancedcnnforimagedenoising |
_version_ |
1724169538463858688 |