Research map / Image generation / Image restoration
Diffusion priors for denoising and deblurring
Use diffusion models as priors for blind denoising and real-world deblurring.
41 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Image restoration, keyword cluster “image denoising · blur · noise”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
14 papers in 2023–24 and 17 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 4.5% to 3.9%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Blind Image Deconvolution by Automatic Gradient Activation (CVPR 2016 · 87 citations)
- 2019 · Spatially Variant Linear Representation Models for Joint Filtering (CVPR 2019 · 50 citations)
- 2020 · An Adaptive Neural Network for Unsupervised Mosaic Consistency Analysis in Image Forensics (CVPR 2020 · 61 citations)
- 2021 · Guided Integrated Gradients: An Adaptive Path Method for Removing Noise (CVPR 2021 · 100 citations)
- 2023 · Parallel Diffusion Models of Operator and Image for Blind Inverse Problems (CVPR 2023 · 85 citations)
- 2024 · Unmixing Diffusion for Self-Supervised Hyperspectral Image Denoising (CVPR 2024 · 38 citations)
- 2025 · Deconstructing Denoising Diffusion Models for Self-Supervised Learning (ICLR 2025 · 17 citations)
- 2026 · Self-Supervised One-Step Diffusion Refinement for Snapshot Compressive Imaging (AAAI 2026 · 1 citation)
Earliest papers here
- Blind Image Deconvolution by Automatic Gradient Activation (CVPR 2016 · 87 citations)
- Random Mixed Field Model for Mixed-Attribute Data Restoration (AAAI 2016 · 3 citations)
- Spatially Variant Linear Representation Models for Joint Filtering (CVPR 2019 · 50 citations)
What has been tried
31 papers here report 342 measured changes (a component added, removed or swapped, compared with a baseline): helped 71%, no clear effect 8%, hurt 17%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add matched Markov-chain start-state selection | KAIST · inference time | 311s -> 41s | Helped | Unmixing Diffusion for Self-Supervised Hyperspectral Image Denoising (CVPR 2024 · 38 citations) |
| remove class-conditioning | ImageNet · linear probe accuracy | 57.5% to 62.5% | Helped | Deconstructing Denoising Diffusion Models for Self-Supervised Learning (ICLR 2025 · 17 citations) |
| Add the autoencoding regularization term (λ=1) rather than disable it (λ=0) | MRI 4x and CT 18 views · PSNR (dB) | Helped | Image Reconstruction Via Autoencoding Sequential Deep Image Prior (NeurIPS 2024 · 9 citations) | |
| add the hyper-network NM and customized parameters | SynData simple meshes · Ea | 4.91 -> 4.32 | Helped | Hyper-MD: Mesh Denoising with Customized Parameters Aware of Noise Intensity and Geometric Characteristics (CVPR 2024 · 6 citations) |
| Add LKPN-guided EAC structural conditioning to ControlNet | Real Images · MANIQA | 0.5544 -> 0.5963 | Helped | DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · 5 citations) |
| Add RG-Diff and SAM prompt fusion | SIDD · PSNR | 51.32 -> 51.55 | Helped | Prompt-SID: Learning Structural Representation Prompt via Latent Diffusion for Single Image Denoising (AAAI 2025 · 4 citations) |
| Add adaptive likelihood estimation (ALE) | SIDD Validation · PSNR/SSIM | 32.12/0.741 -> 34.63/0.870 | Helped | Diffusion Priors for Variational Likelihood Estimation and Image Denoising (NeurIPS 2024 · 4 citations) |
| Use an anechoic clean-speech prior instead of a reverberant-speech prior | SMS-WSJ 3-channel · SDR (dB) | 15.1±1.1 -> 15.8±1.2 | Helped | ArrayDPS: Unsupervised Blind Speech Separation with a Diffusion Prior (ICML 2025 · 0 citations) |
| Use disentangled decoders rather than a shared single decoder during training | Super resolution · PSNR | approximately over a 1 dB to 1.5 dB improvement | Helped | UGoDIT: Unsupervised Group Deep Image Prior Via Transferable Weights (NeurIPS 2025 · 0 citations) |
| Add detail refinement | M4Raw test, T1 · PSNR | 32.59 -> 32.77 | Helped | Score-based Self-supervised MRI Denoising (ICLR 2025 · 0 citations) |
| remove the Fusion process | In-vivo dMRI denoising · image drift | Hurt | Self-Supervised Diffusion MRI Denoising via Iterative and Stable Refinement (ICLR 2025 · 0 citations) | |
| Increase the Gaussian-noise lubrication coefficient γ | Natural-image interpolation · image artifacts | Helped | NoiseDiffusion: Correcting Noise for Image Interpolation with Diffusion Models beyond Spherical Linear Interpolation (ICLR 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 9 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- ImageNet · 2 of 9 papers
- fastMRI · 2 of 9 papers
Baselines
- Noisier2Noise · 2 of 9 papers
- Recorrupted2Recorrupted · 2 of 9 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 31 papers; from OpenAlex and the papers' affiliations).
- Michigan State University · 3 papers
- ETH Zurich · 2 papers
- Harbin Institute of Technology · 2 papers
- Stanford University · 2 papers
- University of Illinois, Urbana Champaign · 2 papers
- University of Michigan · 2 papers
Use it
Download BibTeX of 35 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Image deblurring (73 papers)
- Self-supervised image denoising (60 papers)
- Raw and low-light noise modeling (53 papers)
Elsewhere in the topic
- Dehazing and deraining (118 papers)
- Efficient restoration architectures (113 papers)
- Inverse imaging with learned priors (109 papers)
- Image harmonization and color transformation (102 papers)
- Reflection removal (95 papers)
- Pansharpening and spectral fusion (84 papers)
- Low-light image enhancement (72 papers)
- Learned image compression (69 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).