mlsanityData to 7 Oct 2026

Deblurring via Stochastic Refinement

arXiv 2112.02475 ↗First posted 5 Dec 2021cs.CV eess.IVcited by 248

On the map: Diffusion models → Inverse problems and image restoration

Abstract (arXiv)

Image deblurring is an ill-posed problem with multiple plausible solutions for a given input image. However, most existing methods produce a deterministic estimate of the clean image and are trained to minimize pixel-level distortion. These metrics are known to be poorly correlated with human perception, and often lead to unrealistic reconstructions. We present an alternative framework for blind deblurring based on conditional diffusion models. Unlike existing techniques, we train a stochastic sampler that refines the output of a deterministic predictor and is capable of producing a diverse set of plausible reconstructions for a given input. This leads to a significant improvement in perceptual quality over existing state-of-the-art methods across multiple standard benchmarks. Our predict-and-refine approach also enables much more efficient sampling compared to typical diffusion models. Combined with a carefully tuned network architecture and inference procedure, our method is competitive in terms of distortion metrics such as PSNR. These results show clear benefits of our diffusion-based method for deblurring and challenge the widely used strategy of producing a single, deterministic reconstruction.

Key facts. Cited by 248 papers in the map (data through 7 Oct 2026). Placed in Inverse problems and image restoration, one of 14 directions in Diffusion models. Builds on 52 of its 77 references that are in the map. First posted on arXiv 5 Dec 2021.

Builds on

52 of 77 references.

Closest ancestors · 1

Same direction, newest first

  1. 2021SRDiff: Single Image Super-Resolution with Diffusion Probabilistic Models30 Apr

Other references · 51

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  6. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,609
  7. 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,543
  8. 2021Image Super-Resolution via Iterative Refinementcited by 1,506
  9. 2018Demystifying MMD GANscited by 1,268
  10. 2016Instance Normalization: The Missing Ingredient for Fast Stylizationcited by 1,187
  11. 2021Multi-Stage Progressive Image Restorationcited by 813
  12. 2016Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurringcited by 722
38 more
  1. 2020WaveGrad: Estimating Gradients for Waveform Generationcited by 620
  2. 2017The Perception-Distortion Tradeoffcited by 582
  3. 2017DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networkscited by 366
  4. 2019DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Bettercited by 361
  5. 2017Compressed Sensing using Generative Modelscited by 337
  6. 2018Scale-recurrent Network for Deep Image Deblurringcited by 332
  7. 2020High-Fidelity Generative Image Compressioncited by 331
  8. 2021Rethinking Coarse-to-Fine Approach in Single Image Deblurringcited by 324
  9. 2020Deep Learning Techniques for Inverse Problems in Imagingcited by 290
  10. 2021Robust Compressed Sensing MRI with Deep Generative Priorscited by 284
  11. 2021HINet: Half Instance Normalization Network for Image Restorationcited by 236
  12. 2021Gotta Go Fast When Generating Data with Score-Based Modelscited by 218
  13. 2020SRFlow: Learning the Super-Resolution Space with Normalizing Flowcited by 209
  14. 2019Deep Stacked Hierarchical Multi-patch Network for Image Deblurringcited by 205
  15. 2021SNIPS: Solving Noisy Inverse Problems Stochasticallycited by 191
  16. 2020Human-Aware Motion Deblurringcited by 174
  17. 2021On Fast Sampling of Diffusion Probabilistic Modelscited by 170
  18. 2015Learning a Convolutional Neural Network for Non-uniform Motion Blur Removalcited by 148
  19. 2018The Contextual Loss for Image Transformation with Non-Aligned Datacited by 135
  20. 2021Learning to Efficiently Sample from Diffusion Probabilistic Modelscited by 129
  21. 2020Spatially-Attentive Patch-Hierarchical Network for Adaptive Motion Deblurringcited by 119
  22. 2018Distribution Matching Losses Can Hallucinate Features in Medical Image Translationcited by 99
  23. 2021Noise Estimation for Generative Diffusion Modelscited by 94
  24. 2016A Neural Approach to Blind Motion Deblurringcited by 69
  25. 2021WaveGrad 2: Iterative Refinement for Text-to-Speech Synthesiscited by 65
  26. 2021A Theory of the Distortion-Perception Tradeoff in Wasserstein Spacecited by 52
  27. 2020Composing Normalizing Flows for Inverse Problemscited by 48
  28. 2021Stochastic Image Denoising by Sampling from the Posterior Distributioncited by 45
  29. 2021BANet: Blur-aware Attention Networks for Dynamic Scene Deblurringcited by 44
  30. 2018Blind Image Deconvolution using Deep Generative Priorscited by 40
  31. 2017Learning Blind Motion Deblurringcited by 36
  32. 2020Robust Compressed Sensing using Generative Modelscited by 33
  33. 2021High Perceptual Quality Image Denoising with a Posterior Sampling CGANcited by 33
  34. 2021Fairness for Image Generation with Uncertain Sensitive Attributescited by 31
  35. 2020Projected Distribution Loss for Image Enhancementcited by 29
  36. 2018Maintaining Natural Image Statistics with the Contextual Losscited by 19
  37. 2017Deep Generative Filter for Motion Deblurringcited by 14
  38. 2015Removing Camera Shake via Weighted Fourier Burst Accumulationcited by 13

50 most cited of 51 All 51 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 31; 2023: 80; 2024: 68; 2025: 45; 2026: 24.

248 papers.

Sort:

By area:

  1. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  2. 2022Video Diffusion Modelscited by 2,280
  3. 2022Imagen Video: High Definition Video Generation with Diffusion Modelscited by 1,765
  4. 2022Diffusion Models: A Comprehensive Survey of Methods and Applicationscited by 1,107
  5. 2022Denoising Diffusion Restoration Modelscited by 951
  6. 2022eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoiserscited by 921
  7. 2022Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Modelcited by 574
  8. 2022MaxViT: Multi-Axis Vision Transformercited by 476
  9. 2023Universal Guidance for Diffusion Modelscited by 426
  10. 2023Understanding Diffusion Objectives as the ELBO with Simple Data Augmentationcited by 293
  11. 2023FreeDoM: Training-Free Energy-Guided Conditional Diffusion Modelcited by 290
  12. 2022Cold Diffusion: Inverting Arbitrary Image Transforms Without Noisecited by 275
38 more
  1. 2022A Survey on Generative Diffusion Modelcited by 238
  2. 2022Restoring Vision in Adverse Weather Conditions with Patch-Based Denoising Diffusion Modelscited by 215
  3. 2023I2SB: Image-to-Image Schr"odinger Bridgecited by 214
  4. 2022MedSegDiff: Medical Image Segmentation with Diffusion Probabilistic Modelcited by 180
  5. 2024A Survey on Diffusion Models for Inverse Problemscited by 173
  6. 2022Generative Modelling With Inverse Heat Dissipationcited by 164
  7. 2023Inversion by Direct Iteration: An Alternative to Denoising Diffusion for Image Restorationcited by 164
  8. 2022GENIE: Higher-Order Denoising Diffusion Solverscited by 127
  9. 2023Consistent View Synthesis with Pose-Guided Diffusion Modelscited by 117
  10. 2023SyncDiffusion: Coherent Montage via Synchronized Joint Diffusionscited by 108
  11. 2023Diffusion Models for Image Restoration and Enhancement: A Comprehensive Surveycited by 108
  12. 2022StoRM: A Diffusion-based Stochastic Regeneration Model for Speech Enhancement and Dereverberationcited by 105
  13. 2022Soft Diffusion: Score Matching for General Corruptionscited by 101
  14. 2023AutoDIR: Automatic All-in-One Image Restoration with Latent Diffusioncited by 98
  15. 2023Global Structure-Aware Diffusion Process for Low-Light Image Enhancementcited by 92
  16. 2022Blurring Diffusion Modelscited by 89
  17. 2023Hierarchical Integration Diffusion Model for Realistic Image Deblurringcited by 87
  18. 2023Refusion: Enabling Large-Size Realistic Image Restoration with Latent-Space Diffusion Modelscited by 86
  19. 2023Wavelet-based Fourier Information Interaction with Frequency Diffusion Adjustment for Underwater Image Restorationcited by 86
  20. 2023HumanMAC: Masked Motion Completion for Human Motion Predictioncited by 81
  21. 2022VIDM: Video Implicit Diffusion Modelscited by 78
  22. 2022Multiscale Structure Guided Diffusion for Image Deblurringcited by 76
  23. 2022DOLCE: A Model-Based Probabilistic Diffusion Framework for Limited-Angle CT Reconstructioncited by 73
  24. 2024Diffusion Models in Low-Level Vision: A Surveycited by 73
  25. 2024An Overview of Diffusion Models: Applications, Guided Generation, Statistical Rates and Optimizationcited by 72
  26. 2023Image Denoising: The Deep Learning Revolution and Beyond -- A Survey Paper --cited by 71
  27. 2023Prompt-tuning latent diffusion models for inverse problemscited by 70
  28. 2023Residual Denoising Diffusion Modelscited by 68
  29. 2023Pyramid Diffusion Models For Low-light Image Enhancementcited by 67
  30. 2023ConsistNet: Enforcing 3D Consistency for Multi-view Images Diffusioncited by 65
  31. 2024Efficient Diffusion Model for Image Restoration by Residual Shiftingcited by 64
  32. 2024DreamClear: High-Capacity Real-World Image Restoration with Privacy-Safe Dataset Curationcited by 64
  33. 2024Diffusion Models, Image Super-Resolution And Everything: A Surveycited by 63
  34. 2023Towards Coherent Image Inpainting Using Denoising Diffusion Implicit Modelscited by 61
  35. 2024DMPlug: A Plug-in Method for Solving Inverse Problems with Diffusion Modelscited by 60
  36. 2023ResDiff: Combining CNN and Diffusion Model for Image Super-Resolutioncited by 53
  37. 2024DarkIR: Robust Low-Light Image Restorationcited by 48
  38. 2024Diff-Plugin: Revitalizing Details for Diffusion-based Low-level Taskscited by 47

50 most cited of 248 All 248 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Unpaired Deblurring via Decoupled Diffusion Model
  2. 2022Progressive Deblurring of Diffusion Models for Coarse-to-Fine Image Synthesis
  3. 2024DIVD: Deblurring with Improved Video Diffusion Model
  4. 2025Plug-and-Play Posterior Sampling for Blind Inverse Problems
  5. 2025DeblurDiff: Real-World Image Deblurring with Generative Diffusion Models
  6. 2023Diffusion with Forward Models: Solving Stochastic Inverse Problems Without Direct Supervision
  7. 2022DriftRec: Adapting diffusion models to blind JPEG restoration
  8. 2023DiracDiffusion: Denoising and Incremental Reconstruction with Assured Data-Consistency