Research map / Image generation / Image restoration
Inverse imaging with learned priors
Solve compressive and inverse imaging problems with unrolled networks and generative priors.
109 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Image restoration, keyword cluster “removal · enhancement · color”
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.
21 papers in 2023–24 and 41 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 6.8% to 9.4%.
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 · Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · 2,321 citations)
- 2017 · An Artificial Agent for Robust Image Registration (AAAI 2017 · 184 citations)
- 2018 · ISTA-Net: Interpretable Optimization-Inspired Deep Network for Image Compressive Sensing (CVPR 2018 · 1,329 citations)
- 2019 · Revisiting Perspective Information for Efficient Crowd Counting (CVPR 2019 · 260 citations)
- 2020 · Fast Symmetric Diffeomorphic Image Registration with Convolutional Neural Networks (CVPR 2020 · 252 citations)
- 2021 · See Through Gradients: Image Batch Recovery via GradInversion (CVPR 2021 · 438 citations)
- 2022 · Robust Invertible Image Steganography (CVPR 2022 · 174 citations)
- 2023 · WIRE: Wavelet Implicit Neural Representations (CVPR 2023 · 172 citations)
- 2024 · Atlantis: Enabling Underwater Depth Estimation with Stable Diffusion (CVPR 2024 · 36 citations)
- 2025 · Using Powerful Prior Knowledge of Diffusion Model in Deep Unfolding Networks for Image Compressive Sensing (CVPR 2025 · 9 citations)
- 2026 · cryoSENSE: Compressive Sensing Enables High-throughput Microscopy with Sparse and Generative Priors on the Protein Cryo-EM Image Manifold (CVPR 2026 · 1 citation)
Earliest papers here
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · 2,321 citations)
- ReconNet: Non-Iterative Reconstruction of Images From Compressively Sensed Measurements (CVPR 2016 · 743 citations)
- An Artificial Agent for Robust Image Registration (AAAI 2017 · 184 citations)
What has been tried
81 papers here report 799 measured changes (a component added, removed or swapped, compared with a baseline): helped 73%, no clear effect 6%, hurt 18%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add the counting decoder during training | JHU-Crowd++ · MAE | 52.29 -> 48.56 | Helped | CrowdDiff: Multi-hypothesis Crowd Density Estimation using Diffusion Models (CVPR 2024 · 35 citations) |
| increase reconstruction stages from 3 to 8 | Set14, CS ratio 0.25 · PSNR | 33.65 -> 34.41 | Helped | CPP-Net: Embracing Multi-Scale Feature Fusion into Deep Unfolding CP-PPA Network for Compressive Sensing (CVPR 2024 · 33 citations) |
| Add random rotations and reflections to the DnCNN denoiser at each iteration | MRI ×8 · PSNR | div. -> 24.6 ± 2.9 | Helped | Equivariant Plug-and-Play Image Reconstruction (CVPR 2024 · 27 citations) |
| add iterative inference to RP-Net | FLARE · Dice (%) | +4.66 | Helped | IIRP-Net: Iterative Inference Residual Pyramid Network for Enhanced Image Registration (CVPR 2024 · 21 citations) |
| remove Gradient Descent from the DMP step | Set11, Set14, and Urban100 at CS ratio = 10% · PSNR | 33.22/0.9277; 31.71/0.8664; 31.39/0.9053 -> 29.89/0.8713; 28.77/0.8015; 27.27/0.8225 | Hurt | Using Powerful Prior Knowledge of Diffusion Model in Deep Unfolding Networks for Image Compressive Sensing (CVPR 2025 · 9 citations) |
| increase the iterative phases from 3 to 7 | Set11 (CS ratio 0.04) · PSNR | 26.62 -> 27.23 | Helped | HUNet: Homotopy Unfolding Network for Image Compressive Sensing (CVPR 2025 · 8 citations) |
| Replace uniform sampling with innovation-based adaptive sampling | BSD68 and Urban100, SR 0.10 and 0.25 · PSNR (dB) | BSD68 SR=0.10 28.38 to 29.54; Urban100 SR=0.25 33.11 to 35.15 | Helped | Sampling Innovation-Based Adaptive Compressive Sensing (CVPR 2025 · 6 citations) |
| Add FDRM to the MR-CCSNet+ configuration | 25% sampling · PSNR (dB) | 33.96 -> 34.37 | Helped | Multi-Cross Sampling and Frequency-Division Reconstruction for Image Compressed Sensing (AAAI 2024 · 5 citations) |
| replace rank-one factorizations with rank-5 factorizations | Oil Painting inpainting (SR=0.1) · PSNR | 37.601 -> 36.837 | Hurt | Deep Rank-One Tensor Functional Factorization for Multi-Dimensional Data Recovery (AAAI 2025 · 4 citations) |
| Increase correction-function size from EDM-T to EDM-H | CIFAR-10 · FID | 2.03→1.93 | Helped | Residual Learning in Diffusion Models (CVPR 2024 · 3 citations) |
| Add diffusion modeling | IEEE · PESQ | 2.21 -> 2.43 | Helped | Complex-Cycle-Consistent Diffusion Model for Monaural Speech Enhancement (AAAI 2025 · 2 citations) |
| fine-tune the denoiser with DEQ | OASIS-1 · Neg. JD % | 0.0567 (0.002) -> 0.0495 (0.001) | Helped | A Plug-and-Play Image Registration Network (ICLR 2024 · 2 citations) |
What reviewers will expect
Datasets and baselines named in the 19 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- fastMRI knee · 3 of 19 papers
- DIV2K · 2 of 19 papers
- Set11 · 2 of 19 papers
- BSD68 · 2 of 19 papers
- Urban100 · 2 of 19 papers
- FastMRI+ · 2 of 19 papers
Baselines
- TransCS · 3 of 19 papers
- OCTUF · 3 of 19 papers
- ScoreMRI · 2 of 19 papers
- DGU-Net+ · 2 of 19 papers
- DPC-DUN · 2 of 19 papers
- CSformer · 2 of 19 papers
- UFC-Net · 2 of 19 papers
- CASNet · 2 of 19 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 62 papers; from OpenAlex and the papers' affiliations).
- Xi’an Jiaotong University · 2 papers
- Boston University · 2 papers
- University of Science and Technology of China · 2 papers
- Northwestern Polytechnical University · 2 papers
- Fudan University · 2 papers
Use it
Download BibTeX of 40 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
- Dehazing and deraining (118 papers)
- Image harmonization and color transformation (102 papers)
- Reflection removal (95 papers)
- Pansharpening and spectral fusion (84 papers)
- Low-light image enhancement (72 papers)
Elsewhere in the topic
- Efficient restoration architectures (113 papers)
- Image deblurring (73 papers)
- Learned image compression (69 papers)
- Priors for restoration networks (66 papers)
- Diffusion-based universal restoration (61 papers)
- Self-supervised image denoising (60 papers)
- Raw and low-light noise modeling (53 papers)
- Diffusion priors for denoising and deblurring (41 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).