Research map / Image generation / Image restoration
Image harmonization and color transformation
Harmonize composites and decompose illumination using pixel-wise color transforms.
102 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.
27 papers in 2023–24 and 31 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 8.7% to 7.1%.
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 · Automatic Image Cropping : A Computational Complexity Study (CVPR 2016 · 124 citations)
- 2017 · Radiometric Calibration From Faces in Images (CVPR 2017 · 10 citations)
- 2018 · Image Inpainting via Generative Multi-column Convolutional Neural Networks (NeurIPS 2018 · 254 citations)
- 2019 · ESIR: End-To-End Scene Text Recognition via Iterative Image Rectification (CVPR 2019 · 332 citations)
- 2020 · Rethinking the Image Fusion: A Fast Unified Image Fusion Network based on Proportional Maintenance of Gradient and Intensity (AAAI 2020 · 594 citations)
- 2021 · Invertible Image Signal Processing (CVPR 2021 · 129 citations)
- 2022 · RFNet: Unsupervised Network for Mutually Reinforcing Multi-Modal Image Registration and Fusion (CVPR 2022 · 184 citations)
- 2023 · Underwater Ranker: Learn Which Is Better and How to Be Better (AAAI 2023 · 200 citations)
- 2024 · DiffusionLight: Light Probes for Free by Painting a Chrome Ball (CVPR 2024 · 32 citations)
- 2025 · ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency (AAAI 2025 · 5 citations)
- 2026 · DCA-LUT: Deep Chromatic Alignment with 5D LUT for Purple Fringing Removal (AAAI 2026 · 1 citation)
Earliest papers here
- Automatic Image Cropping : A Computational Complexity Study (CVPR 2016 · 124 citations)
- Joint Recovery of Dense Correspondence and Cosegmentation in Two Images (CVPR 2016 · 121 citations)
- Efficient and Robust Color Consistency for Community Photo Collections (CVPR 2016 · 62 citations)
What has been tried
82 papers here report 787 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 4%, hurt 21%, mixed 1%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add iterative inpainting and LoRA to SDXL† | Laval Indoor, three spheres, diffuse · si-RMSE | 0.20 -> 0.14 | Helped | DiffusionLight: Light Probes for Free by Painting a Chrome Ball (CVPR 2024 · 32 citations) |
| calculate distribution distance in the frequency domain instead of feature space | DPED · PSNR | 22.916 -> 23.048 | Helped | Misalignment-Robust Frequency Distribution Loss for Image Transformation (CVPR 2024 · 17 citations) |
| remove the phase branch and phase loss | ZRR · SSIM | 0.7413 -> 0.7546 | Helped | Enhancing RAW-to-sRGB with Decoupled Style Structure in Fourier Domain (AAAI 2024 · 16 citations) |
| Use learnable Gabor wavelet filters instead of SIREN filters | DIV2KW homography (out-of-scale) · mPSNR | 26.81 -> 26.90 | Helped | Towards Progressive Multi-Frequency Representation for Image Warping (CVPR 2024 · 12 citations) |
| Add EFReg camera regression and condition on predicted elevation and focal length during denoising | GSO single-view reconstruction · 3D reconstruction quality | Helped | Era3D: High-Resolution Multiview Diffusion using Efficient Row-wise Attention (NeurIPS 2024 · 10 citations) | |
| Add RAW image condition w | ZRR · structural detail | Helped | DiffRAW: Leveraging Diffusion Model to Generate DSLR-Comparable Perceptual Quality sRGB from Smartphone RAW Images (AAAI 2024 · 7 citations) | |
| scale log RGB to [0, 1] | RAW10 image classification · accuracy | Hurt | Logarithmic Lenses: Exploring Log RGB Data for Image Classification (CVPR 2024 · 7 citations) | |
| Add chromaticity map extraction to the CMA method | ROD HQ→ROD LQ · AUC | a reduction of 0.0952 in AUC | Helped | CMA: A Chromaticity Map Adapter for Robust Detection of Screen-Recapture Document Images (CVPR 2024 · 6 citations) |
| directly estimate distortion flow from the latent GS image to RS frames | Fastec-RS · PSNR | 28.39 -> 28.88 | Helped | Rolling Shutter Correction with Intermediate Distortion Flow Estimation (CVPR 2024 · 6 citations) |
| decouple grayscale detail reconstruction from color mapping | ZRR test set, Original GT · PSNR | 20.12 -> 20.93 | Helped | ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency (AAAI 2025 · 5 citations) |
| Add Laplacian decomposition | Gehler · mean angular error (°) | 3.71 -> 2.35 | Helped | GCC: Generative Color Constancy via Diffusing a Color Checker (CVPR 2025 · 5 citations) |
| replace token-wise conditioning with channel concatenation | Laval Indoor (three-sphere) · angular error (Diffuse) | 2.56 -> 7.09 | Hurt | LuxDiT: Lighting Estimation with Video Diffusion Transformer (NeurIPS 2025 · 3 citations) |
What reviewers will expect
Datasets and baselines named in the 18 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- PyNet · 2 of 18 papers
- AWNet · 2 of 18 papers
- UNet · 2 of 18 papers
- Restormer · 2 of 18 papers
- Uformer · 2 of 18 papers
- WDiff · 2 of 18 papers
- DiT · 2 of 18 papers
- RealFill · 2 of 18 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 58 papers; from OpenAlex and the papers' affiliations).
- Tencent · 4 papers
- National University of Singapore · 3 papers
- Peng Cheng Laboratory · 3 papers
- Nanyang Technological University · 3 papers
- Tsinghua University · 3 papers
- Tongji University · 3 papers
- Shanghai Jiao Tong University · 3 papers
- University of Electronic Science and Technology of China · 2 papers
- Megvii · 2 papers
- Peking 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)
- Inverse imaging with learned priors (109 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).