Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
Open accessThe DOI matches its Crossref record; 53 of the 53 references it cites were found in a registry or an index.
Abstract
The discriminative model learning for image denoising has been recently attracting considerable attentions due to its favorable denoising performance. In this paper, we take one step forward by investigating the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image denoising. Specifically, residual learning and batch normalization are utilized to speed up the training process as well as boost the denoising performance. Different from the existing discriminative denoising models which usually train a specific model for additive white Gaussian noise at a certain noise level, our DnCNN model is able to handle Gaussian denoising with unknown noise level (i.e., blind Gaussian denoising). With the residual learning strategy, DnCNN implicitly removes the latent clean image in the hidden layers. This property motivates us to train a single DnCNN model to tackle with several general image denoising tasks, such as Gaussian denoising, single image super-resolution, and JPEG image deblocking. Our extensive experiments demonstrate that our DnCNN model can not only exhibit high effectiveness in several general image denoising tasks, but also be efficiently implemented by benefiting from GPU computing.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“The architecture of DnCNN is an end-to-end trainable deep CNN for Gaussian denoising and it adopts the residual learning strategy to remove the latent clean image from noisy observation, as reported in [6] by Zhang et al.”
From Median Filter Aided CNN Based Image Denoising: An Ensemble Approach (2021) · cites this paper as [6]
The cited source says what the claim attributes to it.
In the citing paper
“it has been pointed out that the combination of batch normalization and residual learning is particularly helpful for Gaussian denoising since they are beneficial to each other.”
From Learning Deep CNN Denoiser Prior for Image Restoration (2017) · cites this paper as [65]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 7×
“According to [22], when the original mapping is more like an identity mapping, the residual mapping will be much easier to be optimized.”
BackgroundInfluentialVery Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148 · cited 4×
“The representative achievements in training CNN models include Rectified Linear Unit (ReLU) [20], tradeoff be-tween depth and width [19], [26], parameter initialization [27], gradient-based optimization algorithms [28], [29], [30], batch normalization [21] and residual learning [22].”
BackgroundMethodInfluentialImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155 · cited 2×
“The representative achievements in training CNN models include Rectified Linear Unit (ReLU) [20], tradeoff between depth and width [19], [26], parameter initialization [27], gradient-based optimization algorithms [28], [29], [30], batch normalization [21] and residual learning [22].”
BackgroundAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 2×
“The representative achievements in training CNN models include Rectified Linear Unit (ReLU) [20], tradeoff be-tween depth and width [19], [26], parameter initialization [27], gradient-based optimization algorithms [28], [29], [30], batch normalization [21] and residual learning [22].”
BackgroundMethodGoing deeper with convolutions
Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., & Rabinovich, A. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298594 · Cited by 48,562
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818 · cited 6×
“Batch normalization [21] is proposed to alleviate the internal covariate shift by incorporating a normalization step and a scale and shift step before the nonlinearity in each layer.”
BackgroundMethodInfluentialDelving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.123 · Cited by 19,741
Nonlinear total variation based noise removal algorithms
Rudin, L. I., Osher, S., & Fatemi, E. (1992) · Physica D: Nonlinear Phenomena · doi:10.1016/0167-2789(92)90242-f · Cited by 15,704 · cited 1×
“Over the past few decades, various models have been exploited for modeling image priors, including nonlocal self-similarity (NSS) models [1], [2], [3], [4], sparse models [4], [5], [6], gradient models [7], [8], [9] and Markov random field (MRF) models [10], [11], [12].”
BackgroundBatch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S. (2024) · arXiv (Cornell University) · doi:10.57702/o9raffed · Cited by 14,214
Image denoising by sparse 3-D transform-domain collaborative filtering
Dabov, K., Foi, A., Katkovnik, V., & Egiazarian, K. (2007) · IEEE Transactions on Image Processing · doi:10.1109/tip.2007.901238 · Cited by 9,283 · cited 7×
“, blind Gaussian denoising), DnCNN with a single model can still outperform BM3D [2] and TNRD [16] trained”
BackgroundMethodInfluential
Cited by
Items 1,001–1,020 of 1,020
Chae, S. G., Bae, S. J., & Oh, K.-Y. (2025) · Journal of Energy Storage · doi:10.1016/j.est.2024.114826 · Cited by 47
Residual error based knowledge distillation
Gao, M., Wang, Y., & Wan, L. (2021) · Neurocomputing · doi:10.1016/j.neucom.2020.10.113 · Cited by 45
Convolutional Neural Network with Median Layers for Denoising Salt-and-Pepper Contaminations
Liang, L., Deng, S., Gueguen, L., Wei, M., Wu, X., & Qin, J. (2019) · arXiv · doi:10.48550/arxiv.1908.06452 · Cited by 44
“Image denoising is a well-studied yet not well-solved problem [1], [2], [3], [4], [5], [6], [7], where the goal is to recover the underlying signal from its contaminated observation.”
BackgroundMethodInfluentialRecaptured screen image demoiréing
Yue, H., Mao, Y., Liang, L., Xu, H., Hou, C., & Yang, J. (2021) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.2969984 · Cited by 42
Stochastic Image Denoising by Sampling from the Posterior Distribution
Kawar, B., Vaksman, G., & Elad, M. (2021) · 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · doi:10.1109/iccvw54120.2021.00213 · Cited by 42
“Various successful algorithms, both classically oriented and deep learning based, were proposed over the years for handling this task, such as NLM [13], KSVD [19], BM3D [17], EPLL [59], WNNM [21], TNRD [16], DnCNN [53], NLRN [29] and others [52, 36, 4, 45, 25, 54, 42, 44, 26, 57].”
BackgroundMethodZhang, J., Cao, Y., Wang, Y., Wen, C., & Chen, C. W. (2018) · Proceedings of the 26th ACM international conference on Multimedia · doi:10.1145/3240508.3240653 · Cited by 40
“age processing methods. Naturally, a deeper and larger CNN will have better modeling capacity. The most powerful CNNs for image processing tasks usually have dozens of layers and hundreds of channels [39,28,44], thus resulting in millions of network parameters. This leads to a high computational cost, which limits its…”
BackgroundDistribution-aware Adaptive Multi-bit Quantization
Zhao, S., Yue, T., & Hu, X. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.00916 · Cited by 33
“In the past decades, deep neural networks have achieved great successes in many fields [15, 21, 37, 7].”
BackgroundDIPPAS: a deep image prior PRNU anonymization scheme
Picetti, F., Mandelli, S., Bestagini, P., Lipari, V., & Tubaro, S. (2022) · EURASIP Journal on Information Security · doi:10.1186/s13635-022-00128-7 · Cited by 26
“We compare our proposed strategy with [13] `(3)1 (yellow ), [13] `(5)1 (purple N) and [6] (green ). ing DIPPAS and the methods of [13], [6] and DnCNN [23].”
BackgroundMethodInfluentialMedian Filter Aided CNN Based Image Denoising: An Ensemble Approach
Dey, S., Bhattacharya, R., Schwenker, F., & Sarkar, R. (2021) · Algorithms · doi:10.3390/a14040109 · Cited by 22
MaxDropout: Deep Neural Network Regularization Based on Maximum Output Values
do Santos, C. F. G., Colombo, D., Roder, M., & Papa, J. P. (2021) · 2020 25th International Conference on Pattern Recognition (ICPR) · doi:10.1109/icpr48806.2021.9412733 · Cited by 23
“Since then, several other works [7], [8], [9], including the current state-of-the-art on image classification [10], also highlighted its importance.”
BackgroundA simulation-based few samples learning method for surface defect segmentation
Wei, T., Cao, D., Zheng, C., & Yang, Q. (2020) · Neurocomputing · doi:10.1016/j.neucom.2020.06.090 · Cited by 21
A Bias-Reducing Loss Function for CT Image Denoising
Nagare, M., Melnyk, R., Rahman, O., Sauer, K. D., & Bouman, C. A. (2021) · ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp39728.2021.9413855 · Cited by 20
“To do this, we used the architecture of a CNN denoiser adopted from [16] with a single input channel and 17 convolution layers.”
MethodMoiré pattern removal via attentive fractal network
Xu, D., Chu, Y., & Sun, Q. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00244 · Cited by 19
“The second input image is severely contaminated, and the results of DNCNN, SGN and UNet are over-smoothed with unpleasant artifacts.”
MethodISP Meets Deep Learning: A Survey on Deep Learning Methods for Image Signal Processing
Santos, C. F. G. D., Arrais, R. R., Silva, J. V. S. D., Silva, M. H. M. D., Neto, W. B. G. D. A., Lopes, L. T., Bileki, G. A., Lima, I. O., Rondon, L. B., Souza, B. M. D., et al. (2025) · ACM Computing Surveys · doi:10.1145/3708516 · Cited by 15
“For example, deep learning models can achieve state-of-the-art performance in tasks like denoising [87], demosaicing [46], and super-resolution [133], often surpassing classical algorithms.”
RaScaNet: Learning Tiny Models by Raster-Scanning Images
Yoo, J., Lee, D., Son, C., Jung, S., Yoo, B., Choi, C., Han, J.-J., & Han, B. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.01346 · Cited by 14
“Despite many successes of deep neural networks in computer vision applications [13, 20, 38], it is still challenging to deploy a convolutional neural network (CNN) on an MCU due to the on-chip memory restrictions and the problems related to power consumption and latency [12, 21].”
BackgroundFrom Black Boxes to Glass Boxes: Explainable AI for Trustworthy Deepfake Forensics
Qian, H., Xia, L., Ge, R., Fan, Y., Wang, Q., & Jing, Z. (2025) · Cryptography · doi:10.3390/cryptography9040061 · Cited by 10
“Zhang et al. [18] pioneered this approach with DnCNN, a CNN designed to predict the residual (noise) rather than the clean image.”
C3net: demoiréing network attentive in channel, color, and concatenation
Kim, S., Nam, H., Kim, J., & Jeong, J. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00221 · Cited by 9
“…C3Net with moir´e-patterned images, demoir´ed images from Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising (DnCNN) [43], Image Super-Resolution via Deep Recursive Residual Network (DRRN) [29], Image Super-Resolution Using Very Deep Residual Channel Attention…”
MethodMultibranch Wavelet-Based Network for Image Demoiréing
Yeh, C.-H., Lo, C., & He, C.-H. (2024) · Sensors · doi:10.3390/s24092762 · Cited by 7
“These methods are often able to learn effective feature representations from large amounts of data, enabling them to better handle image restoration problems [22–29].”
BackgroundChen, Z., Lu, L., Yuan, Z., Zhu, Y., Li, Y., Yuan, C., & Deng, W. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i2.27889 · Cited by 3
“…has demonstrated remarkable potential in restoring high-quality facial images from unknown degradations, including low resolution (Dong et al. 2014; Lim et al. 2017), blur (Kupyn et al. 2018; Shen et al. 2018), noise (Zhang et al. 2017), and compression artifacts (Dong et al. 2015; Li et al. 2014).”
BackgroundGlaze: protecting artists from style mimicry by text-to-image models
Shan, S., Cryan, J., Wenger, E., Zheng, H., Hanocka, R., & Zhao, B. Y. (2023) · arXiv · doi:10.48550/arxiv.2302.04222
“For denoising, we ran a state-of-the-art CNN-based image denoiser [104] that is specifically trained to remove “additive Gaussian noise” (the same type of noise added to cloaked artwork).”
Method
Metadata
- Authors
- Zhang, K., Zuo, W., Chen, Y., Meng, D., & Zhang, L.
- Affiliations
- Hong Kong Polytechnic University (HK)
- Harbin Institute of Technology (CN)
- Xi'an Jiaotong University (CN)
- Venue
- IEEE Transactions on Image Processing
- Published
- 2017-07
- Volume
- 26
- Issue
- 7
- Pages
- 3142-3155
- ISSN
- 1057-7149 (ISSN-L)
- 1941-0042
- OpenAlex
- W2508457857
- Cited by
- 8,983
- References
- 52
- Citations by year
- 2026: 889 · 2025: 1243 · 2024: 1129 · 2023: 1153 · 2022: 1123 · 2021: 1187 · 2020: 997 · 2019: 761 · 2018: 399 · 2017: 95
- FWCI
- 255.51
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Transactions on Image Processing (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- PolyU Institutional Research Archive (Hong Kong Polytechnic University) (submittedVersion)
- TUScholarShare (Temple University) (submittedVersion)
- arXiv (Cornell University)
- TUScholarShare (Temple University)
- Funders
- HK RGC GRF (PolyU 5313/13E)
- National Natural Scientific Foundation of China (61671182, 61471146)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image and Signal Denoising Methods
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image Enhancement Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image Processing Techniques
- Keywords
- image denoising, residual learning, batch normalization, unknown noise level, single image super-resolution, GPU acceleration
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.