Research map / Image generation / Image restoration
Perceptual diffusion image compression
Use diffusion decoders to improve perceptual quality at ultra-low bitrates.
28 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2022–2026) · an approach in Image restoration, keyword cluster “compression · distortion · coding”
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.
11 papers in 2023–24 and 16 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 3.5% to 3.7%.
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
- 2022 · The Devil Is in the Details: Window-Based Attention for Image Compression (CVPR 2022 · 298 citations)
- 2023 · Multi-Realism Image Compression With a Conditional Generator (CVPR 2023 · 69 citations)
- 2024 · Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · 27 citations)
- 2025 · Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · 5 citations)
- 2026 · Training-Free Rate-Distortion-Perception Traversal With Diffusion (ICML 2026 · 0 citations)
Earliest papers here
- The Devil Is in the Details: Window-Based Attention for Image Compression (CVPR 2022 · 298 citations)
- Multi-Realism Image Compression With a Conditional Generator (CVPR 2023 · 69 citations)
- Improving Statistical Fidelity for Neural Image Compression with Implicit Local Likelihood Models (ICML 2023 · 7 citations)
What has been tried
20 papers here report 194 measured changes (a component added, removed or swapped, compared with a baseline): helped 68%, no clear effect 6%, hurt 24%, mixed 2%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| replace factorized prior with categorical prior for z | CLIC 2020 test set · BD-Rate | 17.7% | Helped | Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · 27 citations) |
| Apply classifier-free guidance to text conditioning only | MS-COCO 30k · FID | 7.89 -> 4.42; 13.84 -> 4.42 | Helped | Towards image compression with perfect realism at ultra-low bitrates (ICLR 2024 · 8 citations) |
| increase maximum blur from σB,max=0 to σB,max=25 | Kodak · FID | 3.94 -> 3.45 | Helped | Laplacian-guided Entropy Model in Neural Codec with Blur-dissipated Synthesis (CVPR 2024 · 7 citations) |
| Remove diffusion process and auxiliary encoder/loss | Kodak · BDRate (PSNR) | 0 -> +6.91% | Hurt | Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · 5 citations) |
| add global semantic information zh | MSCOCO411 · PSNR | 18.0 -> 20.3; RoI: 17.0 -> 18.9 | Helped | Decouple Distortion from Perception: Region Adaptive Diffusion for Extreme-low Bitrate Perception Image Compression (CVPR 2025 · 4 citations) |
| remove the CCF frequency estimator | LIVE-1 q=10 · PSNR (dB) | 27.76 vs 27.95 | Hurt | JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · 4 citations) |
| Use uniform-noise diffusion instead of Gaussian-noise diffusion | Kodak · reconstruction quality | Helped | Bridging the Gap between Gaussian Diffusion Models and Universal Quantization for Image Compression (CVPR 2025 · 3 citations) | |
| add conditional discriminator only | DIV2K BPG QP 37 · FID | 8.84 -> 8.05 | Helped | Enhancing Quality of Compressed Images by Mitigating Enhancement Bias Towards Compression Domain (CVPR 2024 · 3 citations) |
| add the text adapter to the ELIC encoder | Kodak · LPIPS | more than 4× | Helped | Neural Image Compression with Text-guided Encoding for both Pixel-level and Perceptual Fidelity (ICML 2024 · 2 citations) |
| Add random conditioning | MS-COCO-30k · FID | 18.13 -> 15.46 | Helped | Random Conditioning with Distillation for Data-Efficient Diffusion Model Compression (CVPR 2025 · 1 citation) |
| remove SuperMask | Kodak · BD-rate | +12.45% | Hurt | LotteryCodec: Searching the Implicit Representation in a Random Network for Low-Complexity Image Compression (ICML 2025 · 0 citations) |
| use DDPM + DPS inversion instead of StyleGAN2 + PULSE, StyleGAN2 + ILO, or DDPM + MCG | FFHQ · FID | 15.33 -> 5.347; 26.15 -> 5.347; 135.1 -> 5.347 | Helped | Idempotence and Perceptual Image Compression (ICLR 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 10 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- Kodak · 3 of 10 papers
- CLIC2020 · 3 of 10 papers
- Flickr2W · 2 of 10 papers
- DIV2K · 2 of 10 papers
Baselines
- HiFiC · 5 of 10 papers
- CDC · 4 of 10 papers
- MS-ILLM · 4 of 10 papers
- MRIC · 3 of 10 papers
- VTM · 2 of 10 papers
- ELIC · 2 of 10 papers
- PerCo · 2 of 10 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 27 papers; from OpenAlex and the papers' affiliations).
- Peking University · 2 papers
- Meta · 2 papers
Use it
Download BibTeX of 23 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
- Learned image compression (69 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)
- Image deblurring (73 papers)
- Low-light image enhancement (72 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).