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 901–1,000 of 1,020
Unsupervised Learning-Based Framework for Deepfake Video Detection
Zhang, L., Qiao, T., Xu, M., Zheng, N., & Xie, S. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3182509 · Cited by 64
“Firstly, this CNN network is initialized by us-ing the weights of denoising network in [43].”
MethodInfluentialSelf-citationSimilarity-Informed Self-Learning and Its Application on Seismic Image Denoising
Liu, N., Wang, J., Gao, J., Chang, S., & Lou, Y. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2022.3210217 · Cited by 74
“Among these supervised-learning-based denoising models, the feed-forward denoising convolutional neural networks (DnCNNs) is one of the most typical, proposed by Zhang et al. [51], which consists of convolution, batch normalization (BN), corrected linear unit (ReLU), and residual learning (RL).”
BackgroundRobust Unpaired Image Dehazing via Density and Depth Decomposition
Yang, Y., Wang, C., Guo, X., & Tao, D. (2023) · International Journal of Computer Vision · doi:10.1007/s11263-023-01940-5 · Cited by 74
“…learning capability of deep neural networks (He & Tao, 2020; He et al., 2020), plenty of methods were proposed to solve the image restoration tasks (Zhang et al., 2017; Chang et al., 2020; Zhang et al., 2018; Yang et al., 2020; Ren et al., 2019; Li et al., 2018), as well as image dehazing (Ren et…”
Deep convolutional neural networks to restore single-shot electron microscopy images
Lobato, I., Friedrich, T., & Aert, S. V. (2024) · npj Computational Materials · doi:10.1038/s41524-023-01188-0 · Cited by 74
“Although it has been demonstrated that one architecture developed for a certain image restoration task also performs well for other restoration tasks [60, 63, 58, 64], an architecture for a given task will be data dependent.”
BackgroundAll-optical image denoising using a diffractive visual processor
Işıl, Ç., Gan, T., Ardic, F. O., Mentesoglu, K., Digani, J., Karaca, H., Chen, H., Li, J., Mengü, D., Jarrahi, M., et al. (2024) · Light Science & Applications · doi:10.1038/s41377-024-01385-6 · Cited by 75
“After their training, DNNs generally operate in a non-iterative feed-forward fashion and have shown remarkable performance advantages for image denoising of unknown input images, never seen before [21]–[28].”
Self-citationDeep Image Fingerprint: Towards Low Budget Synthetic Image Detection and Model Lineage Analysis
Sinitsa, S., & Fried, O. (2024) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv57701.2024.00402 · Cited by 75
“The DnCNN-S [17] model was trained separately from the residual extraction procedure.”
BackgroundMethodInfluentialDN-ResNet: Efficient Deep Residual Network for Image Denoising
Ren, H., El‐Khamy, M., & Lee, J. (2019) · Lecture notes in computer science · doi:10.1007/978-3-030-20873-8_14 · Cited by 74
CFSNet: Toward a Controllable Feature Space for Image Restoration
Wang, W., Guo, R., Tian, Y., & Yang, W. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00424 · Cited by 74
“Compared to ARCNN [8] and DnCNN [37], our CFSNet can make a better compromise between artifacts removal and details preservation.”
MethodInfluentialDu, G., Long, S., & Yuan, H. (2020) · IEEE Access · doi:10.1109/access.2020.2964794 · Cited by 74
“We add Gaussian white noise fN (0, 1) to the signals [29].”
BackgroundImproved Low-Count Quantitative PET Reconstruction With an Iterative Neural Network
Lim, H., Chun, I. Y., Dewaraja, Y. K., & Fessler, J. A. (2020) · IEEE Transactions on Medical Imaging · doi:10.1109/tmi.2020.2998480 · Cited by 74
“…an optimization formulation with a mathematical motivation (whereas, for the trained regularizer, [17], [18] brought U-Net [19] and DnCNN that were [20] developed for medical image segmentation and general Gaussian denoising, respectively) and characterized by fewer parameters, thereby avoiding…”
BackgroundWeighted sigmoid gate unit for an activation function of deep neural network
Tanaka, M. (2020) · Pattern Recognition Letters · doi:10.1016/j.patrec.2020.05.017 · Cited by 74
Bao, L., Yang, Z., Wang, S., Bai, D., & Lee, J. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00232 · Cited by 74
“To evaluate the performance of the proposed MCU-Net and MRDN, their SIDD benchmark scores are shown in Table 3 together with the scores of other prior arts, which include BM3D [2], NLM [1], KSVD [34], LPG-PCA [35], FoE [36], MLP [21], WNNM [37], GLIDE [38], TNRD [39], EPLL [40], DnCNN [9], KSVD [41], CycleISP [42] and…”
BackgroundMethodInfluentialWang, J., Deng, X., Xu, M., Chen, C., & Song, Y. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58568-6_24 · Cited by 74
Deep Spatial-Angular Regularization for Light Field Imaging, Denoising, and Super-Resolution
Guo, M., Hou, J., Jin, J., Chen, J., & Chau, L. (2021) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2021.3087485 · Cited by 73
Triply Complementary Priors for Image Restoration
Zha, Z., Wen, B., Yuan, X., Zhou, J. T., Zhou, J., & Zhu, C. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3086049 · Cited by 73
“Most of the existing prior-based image IR algorithms adopted a single image model in their methods [4], [5], [10], [14], [19], [20].”
BackgroundMethodInfluentialLearning RAW-to-sRGB Mappings with Inaccurately Aligned Supervision
Zhang, Z., Wang, H., Liu, M., Wang, R., Zhang, J., & Zuo, W. (2021) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv48922.2021.00431 · Cited by 74
“Motivated by the unprecedented success of deep learning, CNNs have also been investigated to tackle several hard ISP subtasks like image denoising [57, 59, 60], demosaicking [12, 30, 47], auto-white-balance (AWB) [1, 18, 33, 54] and tone mapping [3, 13, 16, 58].”
BackgroundSelf-citationMultiscale Spatial Attention Network for Seismic Data Denoising
Dong, X., Lin, J., Lu, S., Wang, H., & Li, Y. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2022.3178212 · Cited by 73
“In addition, residual learning can also improve the training accuracy, which has been proven by image classification and objective detection [45], [46].”
BackgroundMethodInfluentialFrequency-Domain Deep Guided Image Denoising
Sheng, Z., Liu, X., Cao, S.-Y., Shen, H., & Zhang, H. (2022) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2022.3214375 · Cited by 73
“That is, we replace its backbone from the original modified ADNet (denoted as NEst-A ) with a modified DnCNN [9] (denoted as NEst-D ) which only contains three convolution blocks.”
BackgroundMethodBlind Face Restoration via Integrating Face Shape and Generative Priors
Zhu, F., Zhu, J., Chu, W., Zhang, X., Ji, X., Wang, C., & Tai, Y. Y. (2022) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52688.2022.00751 · Cited by 73
“In recent years, deep convolutional neural networks have gained great success in large amounts of image restoration tasks including denoising [10,55], de-blurring [24, 43], super-resolution [3, 48], inpainting [29, 53], and compression artifacts reduction [5, 9].”
BackgroundAn End-to-End Framework for Joint Denoising and Classification of Hyperspectral Images
Li, X., Ding, M., Gu, Y., & Pižurica, A. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3264587 · Cited by 73
“Deep learning methods for HSI denoising, including [23], [24], and [25], build on the ideas originally developed for traditional grayscale or color image denoising [22].”
BackgroundMethodRandom noise suppression in seismic data: What can deep learning do?
Liu, D., Wang, W., Chen, W., Wang, X., Zhou, Y., & Shi, Z. (2018) · Second International Meeting for Applied Geoscience & Energy · doi:10.1190/segam2018-2998114.1 · Cited by 73
“Zhang et al. (2017) developed denoising convolutional neural networks (DnCNNs) for general image denoising.”
BackgroundPixel super-resolution for lens-free holographic microscopy using deep learning neural networks
Luo, Z., Yurt, A., Stahl, R. J., Lambrechts, A., Reumers, V., Braeken, D., & Lagae, L. (2019) · Optics Express · doi:10.1364/oe.27.013581 · Cited by 73
Ma, R., Hu, H., Xing, S., & Li, Z. (2020) · IEEE Transactions on Image Processing · doi:10.1109/tip.2020.2965294 · Cited by 73
“For image denoising tasks, it has been proved that CNN learns better the image features and performs better in blind Gaussian denoising than noise modeling and prior learning methods [20].”
BackgroundMethodImage Recovery via Transform Learning and Low-Rank Modeling: The Power of Complementary Regularizers
Wen, B., Li, Y., & Bresler, Y. (2020) · IEEE Transactions on Image Processing · doi:10.1109/tip.2020.2980753 · Cited by 73
“…algorithm for a variational formulation, other popular neural networks structures, such as fully connected networks (FCN), convolutional neural networks (CNN), recurrent neural networks (RNN), and U-Net, have been applied to image restoration with state-of-the-art results [26], [41], [44], [45].”
BackgroundRobust Few-Shot Learning for User-Provided Data
Lu, J., Jin, S., Liang, J., & Zhang, C. (2020) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2020.2984710 · Cited by 73
“Some effective denoising methods, such as sparse and redundant representations over learned dictionaries [41], non-local means [42], denoising autoencoder [43], multicolumn stacked sparse denoising autoencoder [44], and denoising convolutional neural networks (CNNs) [45], were developed.”
MethodQi, J., Du, J., Siniscalchi, S. M., Ma, X., & Lee, C. (2020) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2020.2993164 · Cited by 73
“Similarly, auto-encoders with deep learning architectures can achieve better results on image de-noising [13].”
BackgroundDisparity-Aware Domain Adaptation in Stereo Image Restoration
Yan, B., Ma, C., Bare, B., Tan, W., & Hoi, S. C. H. (2020) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr42600.2020.01319 · Cited by 73
“proposed a 20-layer DnCNN [20] to tackle SISR, image denoising, and JPEG deblocking simultaneously.”
BackgroundInfluentialOn the Adversarial Robustness of Visual TransformersPreprint
Shao, R., Shi, Z., Yi, J., Chen, P., & Hsieh, C. (2021) · arXiv (Cornell University) · Cited by 73
DeepZero: Scaling up Zeroth-Order Optimization for Deep Model TrainingPreprint
Chen, A., Zhang, Y., Jia, J., Diffenderfer, J., Liu, J., Parasyris, K., Zhang, Y., Zhang, Z., Kailkhura, B., & Liu, S. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2310.02025 · Cited by 73
“The AE and black-box classifier are DnCNN (Zhang et al., 2017) and ResNet-50, respectively.”
MethodAdaIR: Adaptive All-in-One Image Restoration via Frequency Mining and ModulationPreprint
Cui, Y., Zamir, S. W., Khan, S. H., Knoll, A., Shah, M. A., & Khan, F. S. (2024) · arXiv (Cornell University) · doi:10.48550/arxiv.2403.14614 · Cited by 73
“…deep learning, a great number of convolutional neural network (CNN) based methods have been proposed and attained superior performance over traditional methods on various image restoration tasks, such as image denoising [77,79], dehazing [47,51], deraining [26,50], and motion de-blurring [13, 16].”
Pradhan, B., Al-Najjar, H. A. H., Sameen, M. I., Tsang, I. W.-H., & Alamri, A. M. (2020) · Remote Sensing · doi:10.3390/rs12101676 · Cited by 72
Early Exit or Not: Resource-Efficient Blind Quality Enhancement for Compressed Images
Xing, Q., Xu, M., Li, T., & Guan, Z. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-58517-4_17 · Cited by 72
Color illusions also deceive CNNs for low-level vision tasks: Analysis and implications
Gomez-Villa, A., Martín, A., Vázquez-Corral, J., Bertalmío, M., & Malo, J. (2020) · Vision Research · doi:10.1016/j.visres.2020.07.010 · Cited by 72
Convolutional proximal neural networks and Plug-and-Play algorithms
Hertrich, J., Neumayer, S., & Steidl, G. (2021) · Linear Algebra and its Applications · doi:10.1016/j.laa.2021.09.004 · Cited by 72
“Indeed, it was shown in [48] that we do not have numerical convergence within FBS-PnP for DnCCN [58].”
BackgroundMethodInfluentialFeng, X., Zhang, W., Su, X., & Xu, Z. (2021) · Remote Sensing · doi:10.3390/rs13091858 · Cited by 72
“[32], which combines batch normalization with residual learning and obtained the latest results.”
BackgroundSAE-Net: A Deep Neural Network for SAR Autofocus
Pu, W. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2021.3139914 · Cited by 72
Eficient image denoising using deep learning: A brief survey
Jiang, B., Li, J., Lu, Y., Cai, Q., Song, H., & Lu, G. (2025) · Information Fusion · doi:10.1016/j.inffus.2025.103013 · Cited by 73
Zhussip, M., Soltanayev, S., & Chun, S. Y. (2019) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2019.01050 · Cited by 72
“LDAMP consists of 10 D-AMP layers (iterations) and DnCNN [48] was used in each layer as a denoiser operator.”
BackgroundMethodInfluentialXu, S., Wang, J., Wang, R., Chen, J., & Zou, W. (2019) · Optics Express · doi:10.1364/oe.27.019778 · Cited by 72
Xformer: Hybrid X-Shaped Transformer for Image Denoising
Zhang, J., Zhang, Y., Gu, J., Dong, J., Kong, L., & Yang, X. (2023) · International Conference on Learning Representations · doi:10.48550/arxiv.2303.06440 · Cited by 72
“The CNN-based methods include DnCNN [59], IRCNN [60], FFDNet [61], NLRN [24], MWCNN [25], RNAN [65], RDN [66], DRUNet [58], and P3AN [16].”
BackgroundMethodInfluentialAnagün, Y., Işık, Ş., & Seke, E. (2019) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2019.03.027 · Cited by 71
CIE XYZ Net: Unprocessing Images for Low-Level Computer Vision Tasks
Afifi, M., Abdelhamed, A., Abuolaim, A., Punnappurath, A., & Brown, M. S. (2021) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2021.3070580 · Cited by 71
Recursive Fusion and Deformable Spatiotemporal Attention for Video Compression Artifact Reduction
Zhao, M., Xu, Y., & Zhou, S. (2021) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3474085.3475710 · Cited by 71
“DnCNN [41] is a deep model with batch normalization and residual learning.”
BackgroundMethodInfluentialVideo Super-Resolution via a Spatio-Temporal Alignment Network
Wen, W., Ren, W., Shi, Y., Nie, Y., Zhang, J., & Cao, X. (2022) · IEEE Transactions on Image Processing · doi:10.1109/tip.2022.3146625 · Cited by 71
A Decoupled Multi-Task Network for Shadow Removal
Liu, J., Wang, Q., Fan, H., Li, W., Qu, L., & Tang, Y. (2023) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2023.3252271 · Cited by 71
“In addition, our shadow matte is consistent with the residual image, facilitating the integration of shadow removal into the unified residual learning framework widely used for other image restoration tasks [29], [34], [36], [37], [38], [39], [40], [41], [42].”
MethodRethinking Transformer-Based Blind-Spot Network for Self-Supervised Image Denoising
Li, J., Zhang, Z., & Zuo, W. (2025) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v39i5.32506 · Cited by 71
“The development of learning-based methods (Zhang et al. 2017) has shown superior performance against traditional patch-based ones (Buades, Coll, and Morel 2005; Dabov et al. 2007) on synthetic Gaussian denoising.”
Self-citationKang, J., Kim, S., & Lee, K. M. (2017) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2017.8296236 · Cited by 71
“In image restoration such as superresolution [6, 7, 8], deblurring [9, 10], and denoising [11], CNN already outperforms conventional non-learning based methods.”
MethodFully Convolutional Pixel Adaptive Image Denoiser
Cha, S., & Moon, T. (2019) · IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00426 · Cited by 71
“Comparing methods The baselines we used were: BM3D [7], RED [22], Memnet [31], DnCNN-S and DnCNN-B [38], and N-AIDE [4].”
BackgroundMethodInfluentialDeepInverse: A Python package for solving imaging inverse problems with deep learning
Tachella, J. A., Terris, M., Hurault, S., Wang, A., Davy, L., Scanvic, J., Sechaud, V., Vo, R., Moreau, T., Davies, T., et al. (2025) · The Journal of Open Source Software · doi:10.21105/joss.08923 · Cited by 72
Square Kilometre Array Science Data Challenge 1: analysis and results
Bonaldi, A., An, T., Brüggen, M., Burkutean, S., Coelho, B., Goodarzi, H., Hartley, P., Sandhu, P. K., Wu, C., Yu, L., et al. (2020) · Monthly Notices of the Royal Astronomical Society · doi:10.1093/mnras/staa3023 · Cited by 73
Learning Robust Graph-Convolutional Representations for Point Cloud Denoising
Pistilli, F., Fracastoro, G., Valsesia, D., & Magli, E. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.3047471 · Cited by 70
“At a high level, the block architecture is a residual network that estimates the additive noise component of the input instead of the denoised point cloud, because it has been shown [10] that predicting the residual is easier than directly cleaning the data.”
BackgroundKhmag, A. (2023) · Soft Computing · doi:10.1007/s00500-023-09148-y · Cited by 70
“4: Experimental results of Jar benchmark image represented by PSNR, SSIM and Q-INDEX using several noise removal methods for mixed image (15% impulsive noise) Method Assessments PSNR SSIM Q-index Noisy image 15.55 0.6303 17.09 NL-mean 16.18 0.6462 18.99 WNNM 17.09 0.7039 19.97 DnCNN 19.89 0.7855 20.62 Proposed Method…”
MethodInfluentialBridging the Gap Between Computational Photography and Visual Recognition
VidalMata, R. G., Banerjee, S., RichardWebster, B., Albright, M., Davalos, P., McCloskey, S., Miller, B., Tambo, A. L., Ghosh, S., Nagesh, S., et al. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2020.2996538 · Cited by 70
Denoising of MR images with Rician noise using a wider neural network and noise range division
You, X., Cao, N., Lü, H., Mao, M., & Wanga, W. (2019) · Magnetic Resonance Imaging · doi:10.1016/j.mri.2019.05.042 · Cited by 70
Hyperspectral Image Denoising via Matrix Factorization and Deep Prior Regularization
Lin, B., Tao, X., & Lü, J. (2019) · IEEE Transactions on Image Processing · doi:10.1109/tip.2019.2928627 · Cited by 70
“In our experiment, following [32] and [34], we adopt 400 180×180 gray images for the proposed transfer learning construction scheme.”
BackgroundMethodInfluentialLee, D.-W., Moon, W., & Ye, J. C. (2020) · Nature Machine Intelligence · doi:10.1038/s42256-019-0137-x · Cited by 70
Random Noise Attenuation Based on Residual Convolutional Neural Network in Seismic Datasets
Yang, L., Chen, W., Liu, W., Zha, B., & Zhu, L. (2020) · IEEE Access · doi:10.1109/access.2020.2972464 · Cited by 70
“For the first time in the network, DnCNN [61] has combined residual learning and batch normalization.”
BackgroundMethodInfluentialLiu, Y., Anwar, S., Zheng, L., & Tian, Q. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00262 · Cited by 70
“Given (a) the noisy image, we show the denoising outcome of the bounding box region using DnCNN [40] and GradNet on the top row of (b) and (c), respectively and (d) is the ground truth Gaussian noise.”
BackgroundDensely Self-guided Wavelet Network for Image Denoising
Liu, W., Yan, Q., & Zhao, Y. (2020) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw50498.2020.00224 · Cited by 70
“By involving the strategy of residual learning and adding batch normalization (BN) and ReLU activate function into deep architectures, DnCNN [8] is proposed to handle Gaussian blind denoising and achieves a much higher peak signal to noise ratio (PSNR) than conventional state-of-the-art approaches…”
BackgroundMethodInfluentialA Model-Driven Deep Unfolding Method for JPEG Artifacts Removal
Fu, X., Wang, M., Cao, X., Ding, X., & Zha, Z.-J. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3083504 · Cited by 70
“For image restoration tasks, the most widely used loss function is mean squared error (mse) [23], [46].”
BackgroundMethodInfluentialBrain MRI super-resolution using coupled-projection residual network
Feng, C.-M., Wang, K., Lu, S., Xu, Y. D., & Li, X. (2021) · Neurocomputing · doi:10.1016/j.neucom.2021.01.130 · Cited by 69
GCN-Denoiser: Mesh Denoising with Graph Convolutional Networks
Shen, Y., Fu, H., Du, Z., Chen, X., Burnaev, E. V., Zorin, D., Zhou, K., & Zheng, Y. (2022) · ACM Transactions on Graphics · doi:10.1145/3480168 · Cited by 70
“In contrast, learning-based methods do not make specific assumptions about the underlying features or noise patterns, and have been successfully applied to image denoising [Agostinelli et al. 2013; Burger et al. 2012; Jin et al. 2017; Sterzentsenko et al. 2019; Xie et al. 2012; Yan et al. 2018; Zhang et al. 2017].”
BackgroundMethodDual Convolutional Neural Networks for Low-Level Vision
Pan, J., Sun, D., Zhang, J., Tang, J., Yang, J. J., Tai, Y., & Yang, M. (2022) · International Journal of Computer Vision · doi:10.1007/s11263-022-01583-y · Cited by 69
“…the proposed DualCNN model against the state-of-the-art denoising meth-ods based on statistical priors (BM3D (Dabov et al. 2007), EPLL (Zoran and Weiss 2011), CSF (Schmidt and Roth 2014)) and deep neural networks (MLP (Burger et al. 2012), DNCNN (Zhang et al. 2017b), IRCNN (Zhang et al. 2017c)).”
BackgroundMethodInfluentialVulnerable Road Users and Connected Autonomous Vehicles Interaction: A Survey
Reyes, A., & Ibáñez, J. A. G. (2022) · Sensors · doi:10.3390/s22124614 · Cited by 69
Provable Probabilistic Imaging Using Score-Based Generative Priors
Sun, Y., Wu, Z., Chen, Y., Feng, B. T., & Bouman, K. L. (2024) · IEEE Transactions on Computational Imaging · doi:10.1109/tci.2024.3449114 · Cited by 69
“In the test, PnP and RED are equipped with the pre-trained DnCNN denoisers from [68], while PnP-ULA and DPS use the score networks.”
MethodSelf-citationMulti-Exposure Fusion with CNN Features
Li, H., & Zhang, L. (2018) · Proceedings - International Conference on Image Processing · doi:10.1109/icip.2018.8451689 · Cited by 69
“Network Type Denoising [15] Super-resolution [16] Classification [14] Feature Layer 1 3 10 18 1 3 10 18 1 3 10 18 MEF-SSIM 0.”
BackgroundMethodSelf-citationLin, H., Wei, C., Wang, G., Chen, H., Lin, L., Ni, M., Chen, J., & Zhuo, S. (2019) · Journal of Biophotonics · doi:10.1002/jbio.201800435 · Cited by 69
Liu, J., Koçak, M., Supanich, M. P., & Deng, J. (2020) · Magnetic Resonance Imaging · doi:10.1016/j.mri.2020.05.002 · Cited by 69
Total Deep Variation: A Stable Regularization Method for Inverse Problems
Kobler, E., Effland, A., Kunisch, K. K., & Pock, T. (2021) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2021.3124086 · Cited by 69
“Dataset Scale MemNet [55] VDSR [56] DnCNN-3 [41] DRRN [57] OISR-LF-s [52] TDV(3)3”
BackgroundOn the Adversarial Robustness of Vision Transformers
Shao, R., Shi, Z., Yi, J., Chen, P.-Y., & Hsieh, C.-J. (2021) · arXiv · doi:10.48550/arxiv.2103.15670 · Cited by 68
“(2020) to train a DnCNN (Zhang et al., 2017) denoiser Dθ for each pre-trained model f with stability objective LStab: LStab = E(xi,yi)∈D,δLCE(f(Dθ(xi + δ)), f(xi)) (4) where δ ∼ N (0, σ2I) follows Gaussian distribution.”
Deep spectral CNN for laser induced breakdown spectroscopy
Castorena, J., Oyen, D., Ollila, A., Legett, C., & Lanza, N. (2021) · Spectrochimica Acta Part B Atomic Spectroscopy · doi:10.1016/j.sab.2021.106125 · Cited by 68
“A similar architecture than the one used for pre-processing here was proposed in the context of image denoising in [13] and resulted in state of the art performance over sophisticated handcrafted methods such as the BM3D [14] and denoising by soft-thresholding in a wavelet decomposition [15].”
Feature-Guided CNN for Denoising Images From Portable Ultrasound Devices
Dong, G., Ma, Y., & Basu, A. (2021) · IEEE Access · doi:10.1109/access.2021.3059003 · Cited by 68
“In addition to FFDNet, Denoising Convolutional Neural Networks (DnCNN) [11] is another popular CNN architecture for noise removal.”
MethodDi, Y., & Wang, E. (2021) · Rock Mechanics and Rock Engineering · doi:10.1007/s00603-020-02314-w · Cited by 68
A Survey on Deep Learning Based Channel Estimation in Doubly Dispersive Environments
Gizzini, A. K., & Chafii, M. (2022) · IEEE Access · doi:10.1109/access.2022.3188111 · Cited by 68
Optical and SAR Image Registration Based on Feature Decoupling Network
Xiang, D., Xie, Y., Cheng, J., Xu, Y., Zhang, H., & Zheng, Y. (2022) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2022.3211858 · Cited by 68
Combining Low-Rank and Deep Plug-and-Play Priors for Snapshot Compressive Imaging
Chen, Y., Gui, X., Zeng, J., Zhao, X.-L., & He, W. P. (2023) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2023.3294262 · Cited by 68
“Many popular pre-trained deep denoisers can be selected to solve the denoising problem, such as DnCNN [54] and FFDNet [55].”
BackgroundSelf-citationBC-QNet: A quantum-infused ELM model for breast cancer diagnosis
Bilal, A., Imran, A., Liu, X., Liu, X., Ahmad, Z., Shafiq, M., El‐Sherbeeny, A. M., & Long, H. (2024) · Computers in Biology and Medicine · doi:10.1016/j.compbiomed.2024.108483 · Cited by 68
FlowIE: Efficient Image Enhancement via Rectified Flow
Zhu, Y., Zhao, W., Li, A., Tang, Y., Zhou, J., & Lu, J. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.00010 · Cited by 68
Venkatesh, S. K., Ramachandra, R., Raja, K. B., Spreeuwers, L., Veldhuis, R. N. J., & Busch, C. (2020) · IEEE Winter Conference on Applications of Computer Vision · doi:10.1109/wacv45572.2020.9093488 · Cited by 68
“In the light of the recent advancements in deep learning, we propose to aggregate the denoising approaches [26, 27, 28, 29] using a deep Multiscale Context Aggregation Network (MS-CAN) such that the noise in the morphed image can be easily determined, i.”
MethodDeep‐learning‐based direct inversion for material decomposition
Gong, H., Tao, S., Rajendran, K., Zhou, W., McCollough, C. H., & Leng, S. (2020) · Medical Physics · doi:10.1002/mp.14523 · Cited by 68
Learning Medical Image Denoising with Deep Dynamic Residual Attention Network
Sharif, S. M. A., Naqvi, R. A., & Biswas, M. (2020) · Mathematics · doi:10.3390/math8122192 · Cited by 69
“Nevertheless, to make the comparison as fair as possible, both learning-based methods [12,16] have been trained and tested with the hyperparameters suggested in the actual implementations.”
BackgroundMethodInfluentialMulti-Scale Wavelet Domain Residual Learning for Limited-Angle CT ReconstructionPreprint
Gu, J., & Ye, J. C. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.01382 · Cited by 68
“Recently, deep learning algorithms using convolutional neural network (CNN) show successful results in computer vision applications including image classification [8], denoising [9] and segmentation [10].”
Plug in estimation in high dimensional linear inverse problems a rigorous analysis *
Fletcher, A. K., Pandit, P., Rangan, S., Sarkar, S., & Schniter, P. (2019) · Journal of Statistical Mechanics Theory and Experiment · doi:10.1088/1742-5468/ab321a · Cited by 67
“In this case, [6, 7] established that, whenever the SE has a unique fixed , the estimates x k generated by AMP converge to the Bayes optimal estimate of x 0 from y . b As another example, if x is a natural image, for which an analytical prior is lacking, then (3) suggests to choose g ( ) as a sophisticated image-denois…”
MethodInfluentialHalimi, A., Maccarone, A., Lamb, R. A., Buller, G. S., & McLaughlin, S. (2021) · IEEE Transactions on Computational Imaging · doi:10.1109/tci.2021.3111572 · Cited by 67
“As a third guidance (GI3), we considered the state-of-theart learning based DnCNN denoiser [48], also applied to each scale and wavelength r k ,∀k, `.”
MethodBash, S., Johnson, B., Gibbs, W. N., Zhang, T., Shankaranarayanan, A., & Tanenbaum, L. N. (2021) · Clinical Neuroradiology · doi:10.1007/s00062-021-01121-2 · Cited by 67
“DL-based image denoising methods have demonstrated performance advantages over traditional methods of denoising [9, 10] and may be employed to bolster quality of fast acquisition of MR examinations.”
MethodArtificial intelligence with deep learning in nuclear medicine and radiology
Decuyper, M., Maebe, J., Holen, R. V., & Vandenberghe, S. (2021) · EJNMMI Physics · doi:10.1186/s40658-021-00426-y · Cited by 67
“This methodology was first proposed as a general image denoising method [67] and quickly found its way to medical imaging.”
MethodFlexible and Generalized Real Photograph Denoising Exploiting Dual Meta Attention
Ma, R., Li, S., Zhang, B., Fang, L., & Li, Z. (2022) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2022.3170472 · Cited by 67
“It can be observed that in the Kodak24 dataset, we obtained on average 0.26-dB improvements over the recently best method HPDNet [25] and surpassed CBM3D [1], TNRD [6], DnCNN [7], and FFDNet [9] by a large margin.”
InfluentialA Method Noise-Based Convolutional Neural Network Technique for CT Image Denoising
Singh, P., Diwakar, M., Gupta, R., Kumar, S., Chakraborty, A., Bajal, E., Jindal, M., Shetty, D. K., Sharma, J., Dayal, H., et al. (2022) · Electronics · doi:10.3390/electronics11213535 · Cited by 67
Quantum computational infusion in extreme learning machines for early multi-cancer detection
Bilal, A., Shafiq, M., Obidallah, W. J., Alduraywish, Y. A., & Long, H. (2025) · Journal Of Big Data · doi:10.1186/s40537-024-01050-0 · Cited by 67
Visualizing nanoparticle surface dynamics and instabilities enabled by deep denoising
Crozier, P. A., Leibovich, M., Haluai, P., Tan, M., Thomas, A. M., Vincent, J., Mohan, S., Morales, A. M., Kulkarni, S. A., Matteson, D. S., et al. (2025) · Science · doi:10.1126/science.ads2688 · Cited by 67
Pixel2Pixel: A Pixelwise Approach for Zero-Shot Single Image Denoising
Ma, Q., Jiang, J., Zhou, X., Liang, P., Liu, X., & Ma, J. (2025) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2025.3546870 · Cited by 69
“Furthermore, we also compare three non-zero-shot methods, including the supervised method DnCNN [4], the self-supervised method Neighbour2Neighbour (NB2NB) [10], and a pretraining-based single-image denoising method, TTT-MIM [75].”
Deep iterative reconstruction for phase retrieval
Işıl, Ç., Öktem, F. S., & Koç, A. (2019) · Applied Optics · doi:10.1364/ao.58.005422 · Cited by 67
“These modifications help the DNN to better learn the residual between the input and output images [22].”
BackgroundMethodGAN-Based virtual-to-real image translation for urban scene semantic segmentation
Guo, X. J., Wang, Z., Yang, Q., Lv, W., Liu, X., Wu, Q., & Huang, J. (2019) · Neurocomputing · doi:10.1016/j.neucom.2019.01.115 · Cited by 67
Masuyama, Y., Yatabe, K., Koizumi, Y., Oikawa, Y., & Harada, N. (2019) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp.2019.8682744 · Cited by 67
Dual-Stream Multi-Path Recursive Residual Network for JPEG Image Compression Artifacts Reduction
Jin, Z. J., Iqbal, M. Z., Zou, W., Li, X., & Steinbach, E. G. (2020) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.2982174 · Cited by 67
“In this part, the proposed STRRN is quantitatively and qualitatively compared with 9 state-of-the-art algorithms or networks, including non-DL-based methods: SA-DCT [9], ANCE [44], DicTV [10], and TNRD [45]; DL-based methods: ARCNN [13], RED30 [24], DRRN [33], DnCNN-3 [18] and ARN [36].”
BackgroundMethodInfluentialShi, Z., Chen, Y., Gavves, E., Mettes, P., & Snoek, C. G. M. (2021) · IEEE Transactions on Image Processing · doi:10.1109/tip.2021.3106812 · Cited by 67
“Following [54], to train a blind Gaussian denoiser, random Gaussian noise is added to the clean training depth images in NYU Depth V2 and the clean RGB images in BSDS500, with a noise level σ ∈ [ 0 , 55 ] .”
Self-citationMultistage Dual-Attention Guided Fusion Network for Hyperspectral Pansharpening
Guan, P., & Lam, E. Y. M. (2021) · IEEE Transactions on Geoscience and Remote Sensing · doi:10.1109/tgrs.2021.3114552 · Cited by 66
“In particular, a convolutional neural network (CNN) is applied for various image restoration tasks, such as image superresolution [12], denoising [13], and inpainting [14].”
MethodDeep Learning for Handling Kernel/model Uncertainty in Image Deconvolution
Nan, Y., & Ji, H. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00246 · Cited by 66
“[49, 50, 27]), and showed its superior performance over pre-defined image priors in many experiments.”
BackgroundAssessment framework for deepfake detection in real-world situations
Lu, Y., & Ebrahimi, T. (2024) · EURASIP Journal on Image and Video Processing · doi:10.1186/s13640-024-00621-8 · Cited by 60
“The assessment framework takes this situation into consideration and measures the performance of a detector under different brightness conditions.”
BackgroundMulti-scale dynamic feature encoding network for image demoiréing
Cheng, X., Fu, Z., & Yang, J. (2019) · 2019 IEEE/CVF International Conference on Computer Vision Workshop (ICCVW) · doi:10.1109/iccvw.2019.00432 · Cited by 50
“DnCNN [23],MSFE [5] and Sun [15] on the LCDMoire [20] validation set.”
BackgroundMethodResultInfluential
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.