Research map / Image generation / Image restoration
Pansharpening and spectral fusion
Fuse panchromatic, multispectral, and hyperspectral images with spatial-spectral networks.
84 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Image restoration, keyword cluster “removal · enhancement · color”
State History Benchmarks Tried Checklist Preprints Who Export
Where it stands
Leveling off Several papers from 2025–26 still report on 1 tracked benchmark, but the best result has barely moved since 2024.
22 papers in 2023–24 and 40 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic grew, from 7.1% to 9.2%.
What looks settled
- GaoFen2 (PSNR): best up to 2024 49.14, best in 2025–26 49.34 from 2 papers, less than 1% better.
No longer reported by papers here: GaoFen2 (SAM, last in 2024).
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 · Multispectral Images Denoising by Intrinsic Tensor Sparsity Regularization (CVPR 2016 · 264 citations)
- 2017 · Hyper-Laplacian Regularized Unidirectional Low-Rank Tensor Recovery for Multispectral Image Denoising (CVPR 2017 · 224 citations)
- 2018 · Deep, complex, invertible networks for inversion of transmission effects in multimode optical fibres (NeurIPS 2018 · 9 citations)
- 2019 · Attention-Guided Network for Ghost-Free High Dynamic Range Imaging (CVPR 2019 · 312 citations)
- 2020 · UnModNet: Learning to Unwrap a Modulo Image for High Dynamic Range Imaging (NeurIPS 2020 · 11 citations)
- 2021 · SIPSA-Net: Shift-Invariant Pan Sharpening With Moving Object Alignment for Satellite Imagery (CVPR 2021 · 43 citations)
- 2022 · HDNet: High-Resolution Dual-Domain Learning for Spectral Compressive Imaging (CVPR 2022 · 209 citations)
- 2023 · Spectral Enhanced Rectangle Transformer for Hyperspectral Image Denoising (CVPR 2023 · 132 citations)
- 2024 · SPECAT: SPatial-spEctral Cumulative-Attention Transformer for High-Resolution Hyperspectral Image Reconstruction (CVPR 2024 · 44 citations)
- 2025 · HyperFree: A Channel-adaptive and Tuning-free Foundation Model for Hyperspectral Remote Sensing Imagery (CVPR 2025 · 49 citations)
- 2026 · Polarization Uncertainty-Guided Diffusion Model for Color Polarization Image Demosaicking (AAAI 2026 · 3 citations)
Earliest papers here
- Multispectral Images Denoising by Intrinsic Tensor Sparsity Regularization (CVPR 2016 · 264 citations)
- Computational Imaging for VLBI Image Reconstruction (CVPR 2016 · 70 citations)
- Hyper-Laplacian Regularized Unidirectional Low-Rank Tensor Recovery for Multispectral Image Denoising (CVPR 2017 · 224 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
GaoFen2 · PSNR (higher is better)
8 papers, 2022–2025 · small gains only
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 47.3 | Mutual Information-Driven Pan-Sharpening (CVPR 2022 · 118 citations) |
| 2024 | 49.14 | Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation (AAAI 2024 · 16 citations) |
| 2025 | 49.34 | OTPNet: ODE-inspired Tuning-free Proximal Network for Remote Sensing Image Fusion (AAAI 2025 · 2 citations) |
WorldView-II · PSNR (higher is better)
7 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 41.93 | Memory-Augmented Deep Conditional Unfolding Network for Pan-Sharpening (CVPR 2022) |
| 2024 | 42.23 | Revisiting Spatial-Frequency Information Integration from a Hierarchical Perspective for Panchromatic and Multi-Spectral Image Fusion (CVPR 2024 · 34 citations) |
| 2025 | 42.58 | Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening (AAAI 2025 · 3 citations) |
WorldView-III · PSNR (higher is better)
7 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 30.57 | Memory-Augmented Deep Conditional Unfolding Network for Pan-Sharpening (CVPR 2022) |
| 2024 | 30.99 | Frequency-Adaptive Pan-Sharpening with Mixture of Experts (AAAI 2024 · 31 citations) |
| 2025 | 31.03 | Accelerated Diffusion via High-Low Frequency Decomposition for Pan-Sharpening (AAAI 2025 · 3 citations) |
GaoFen2 · SSIM (higher is better)
6 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 0.989 | Mutual Information-Driven Pan-Sharpening (CVPR 2022 · 118 citations) |
| 2024 | 0.991 | Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation (AAAI 2024 · 16 citations) |
| 2025 | 0.992 | OTPNet: ODE-inspired Tuning-free Proximal Network for Remote Sensing Image Fusion (AAAI 2025 · 2 citations) |
GaoFen2 · ERGAS (lower is better)
6 papers, 2022–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 0.548 | Mutual Information-Driven Pan-Sharpening (CVPR 2022 · 118 citations) |
| 2024 | 0.447 | Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation (AAAI 2024 · 16 citations) |
GaoFen2 · SAM (lower is better)
5 papers, 2022–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2022 | 0.0102 | Mutual Information-Driven Pan-Sharpening (CVPR 2022 · 118 citations) |
| 2024 | 0.009 | Progressive High-Frequency Reconstruction for Pan-Sharpening with Implicit Neural Representation (AAAI 2024 · 16 citations) |
What has been tried
60 papers here report 572 measured changes (a component added, removed or swapped, compared with a baseline): helped 77%, no clear effect 3%, hurt 19%, 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 |
|---|---|---|---|---|
| remove the βc encoding branch | LongKou · OA | 93.40 -> 73.18 | Hurt | HyperFree: A Channel-adaptive and Tuning-free Foundation Model for Hyperspectral Remote Sensing Imagery (CVPR 2025 · 49 citations) |
| Add HSA to the spectral-wise Transformer baseline | Optical filter-based HSI, KAIST 10-scene average · PSNR | 34.02 -> 38.87 | Helped | SPECAT: SPatial-spEctral Cumulative-Attention Transformer for High-Resolution Hyperspectral Image Reconstruction (CVPR 2024 · 44 citations) |
| add ground-truth-derived LPR priors | Kalantari’s dataset · PSNR-L | 42.19 -> 50.89 | Helped | Generating Content for HDR Deghosting from Frequency View (CVPR 2024 · 39 citations) |
| Add local Fourier information by partitioning regions | WorldView-II · PSNR | 42.0354 -> 42.2319 | Helped | Revisiting Spatial-Frequency Information Integration from a Hierarchical Perspective for Panchromatic and Multi-Spectral Image Fusion (CVPR 2024 · 34 citations) |
| Use frequency-domain queries instead of a spatial-domain query | WV3 reduced-resolution · PSNR | 38.884 -> 39.345 | Helped | Wavelet-Assisted Multi-Frequency Attention Network for Pansharpening (AAAI 2025 · 33 citations) |
| Add the mask predictor and frequency split | WorldView-II · PSNR | 41.8998 -> 42.0261 | Helped | Frequency-Adaptive Pan-Sharpening with Mixture of Experts (AAAI 2024 · 31 citations) |
| remove self-attention from FEM | Test1200 · PSNR | 32.98 -> 33.21 | Helped | FMRNet: Image Deraining via Frequency Mutual Revision (AAAI 2024 · 24 citations) |
| Add hybrid-supervised contrast constraint instead of using no loss-function search | SICE · VIF | 1.308 -> 1.585 | Helped | Hybrid-Supervised Dual-Search: Leveraging Automatic Learning for Loss-Free Multi-Exposure Image Fusion (AAAI 2024 · 22 citations) |
| Add MCSS instead of using SS2D | LSUI test · PSNR | 23.88 -> 24.47 | Helped | Adaptive Dual-domain Learning for Underwater Image Enhancement (AAAI 2025 · 18 citations) |
| add ARGM | Pavia · PSNR | 41.86 -> 42.33 | Helped | Self-Learning Hyperspectral and Multispectral Image Fusion via Adaptive Residual Guided Subspace Diffusion Model (CVPR 2025 · 18 citations) |
| Add independent modeling of RGB channels | NTIRE2022 HSI dataset · PSNR | 28.14 -> 31.83 | Helped | Hyperspectral Image Reconstruction via Combinatorial Embedding of Cross-Channel Spatio-Spectral Clues (AAAI 2024 · 18 citations) |
| Replace INR with Pixel Shuffle | CAVE ×4 · PSNR | 52.26±4.37 -> 52.47±4.10 | Hurt | Fourier-enhanced Implicit Neural Fusion Network for Multispectral and Hyperspectral Image Fusion (NeurIPS 2024 · 17 citations) |
What reviewers will expect
Datasets and baselines named in the 20 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- GaoFen-2 · 4 of 20 papers
- CAVE · 4 of 20 papers
- Harvard · 3 of 20 papers
- Chikusei · 3 of 20 papers
- WorldView-II · 2 of 20 papers
- WorldView-III · 2 of 20 papers
- WorldView-3 · 2 of 20 papers
- QuickBird · 2 of 20 papers
Baselines
- PanNet · 3 of 20 papers
- GPPNN · 3 of 20 papers
- FusionNet · 3 of 20 papers
- Brovey · 2 of 20 papers
- GS · 2 of 20 papers
- IHS · 2 of 20 papers
- CANNet · 2 of 20 papers
- SSRNet · 2 of 20 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).
- University of Electronic Science and Technology of China · 6 papers
- Harbin Institute of Technology · 3 papers
- Chinese Academy of Sciences · 3 papers
- University of Science and Technology of China · 3 papers
- Xiamen University · 3 papers
- Northwestern Polytechnical University · 3 papers
- Beijing Institute of Technology · 2 papers
- The Hong Kong Polytechnic University · 2 papers
- ETH Zurich · 2 papers
- Alibaba Group · 2 papers
Use it
Download BibTeX of 36 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)
- Image harmonization and color transformation (102 papers)
- Reflection removal (95 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).