Research map / Image generation
Image restoration: research map
2,009 papers on Image restoration in Image generation, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 4 clusters and 16 approaches. The busiest year so far is 2025.
What it is
Recovering a clean image from a damaged one: removing noise, blur, rain, haze or reflections, brightening low-light photos, and compressing images with little visible loss. Learned priors, including diffusion models, now beat hand-designed filters on most of these tasks.
Terms you will see on this page:
- Degradation
- the process (blur, noise, compression) that damaged the image.
- Image prior
- knowledge of what clean images look like, used to guide the restoration.
- Learned compression
- image codecs whose encoder and decoder are neural networks.
Start reading with the most cited papers here:
- Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022 · 4,037 citations)
- Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement (CVPR 2020 · 2,412 citations)
Within Image generation, its share shrank from 14.9% in 2023–24 to 10.7% in 2025–26 (311 → 436 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Image restoration in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Efficient restoration architectures rising: 10% → 15% of the topic (30 → 65 papers)
- Dehazing and deraining rising: 5% → 9% of the topic (17 → 41 papers)
- Learned image compression rising: 3% → 7% of the topic (10 → 29 papers)
- Image deblurring falling: 8% → 2% of the topic (26 → 10 papers)
- Priors for restoration networks falling: 8% → 2% of the topic (24 → 10 papers)
Papers from 2025–2026 that are already cited most:
- HVI: A New Color Space for Low-light Image Enhancement (CVPR 2025 · 212 citations)
- MambaIRv2: Attentive State Space Restoration (CVPR 2025 · 140 citations)
- DarkIR: Robust Low-Light Image Restoration (CVPR 2025 · 96 citations)
- HyperFree: A Channel-adaptive and Tuning-free Foundation Model for Hyperspectral Remote Sensing Imagery (CVPR 2025 · 49 citations)
Approaches and key papers
Enhancement and removal of artifacts · 1,058 papers
Keywords: removal, enhancement, color
Approaches in this cluster:
- Dehazing and deraining (204 papers)
Remove haze and rain using codebook priors, diffusion priors, and unpaired training. - Inverse imaging with learned priors (178 papers)
Solve compressive and inverse imaging problems with unrolled networks and generative priors. - Image harmonization and color transformation (251 papers)
Harmonize composites and decompose illumination using pixel-wise color transforms. - Reflection removal (158 papers)
Separate reflections from transmission using misaligned data, raw inputs, and flash cues. - Pansharpening and spectral fusion (141 papers)
Fuse panchromatic, multispectral, and hyperspectral images with spatial-spectral networks. - Low-light image enhancement (126 papers)
Brighten dark images with unsupervised, contrastive, and diffusion-based methods.
Most cited and most cited since 2024:
- Zero-Reference Deep Curve Estimation for Low-Light Image Enhancement (CVPR 2020 · 2,412 citations)
- Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR 2016 · 2,321 citations)
- HVI: A New Color Space for Low-light Image Enhancement (CVPR 2025 · 212 citations)
- Text-IF: Leveraging Semantic Text Guidance for Degradation-Aware and Interactive Image Fusion (CVPR 2024 · 187 citations)
General image restoration · 442 papers
Keywords: image restoration, degradation, degraded
Approaches in this cluster:
- Efficient restoration architectures (172 papers)
Design efficient, progressive, and frequency-based networks for general image restoration. - Priors for restoration networks (115 papers)
Use implicit priors, unrolling, and pretrained model priors for image restoration. - Diffusion-based universal restoration (98 papers)
Apply generative diffusion priors to unified blind image restoration. - Real-world restoration with unaligned data (57 papers)
Restore images in dynamic scenes, burst, and under-display cameras using pseudo-supervision.
Most cited and most cited since 2024:
- Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022 · 4,037 citations)
- Learning Deep CNN Denoiser Prior for Image Restoration (CVPR 2017 · 2,114 citations)
- Omni-Kernel Network for Image Restoration (AAAI 2024 · 275 citations)
- Wavelet-based Fourier Information Interaction with Frequency Diffusion Adjustment for Underwater Image Restoration (CVPR 2024 · 187 citations)
Denoising and deblurring · 370 papers
Keywords: image denoising, blur, noise
Approaches in this cluster:
- Image deblurring (124 papers)
Remove blur with blind, unsupervised, and kernel-aware deblurring networks. - Self-supervised image denoising (98 papers)
Denoise images without clean targets using blind-spot and Noise2Noise-style training. - Raw and low-light noise modeling (89 papers)
Model camera noise physics to train raw denoisers. - Diffusion priors for denoising and deblurring (59 papers)
Use diffusion models as priors for blind denoising and real-world deblurring.
Most cited and most cited since 2024:
- Noise2Void - Learning Denoising From Single Noisy Images (CVPR 2019 · 1,426 citations)
- Scale-Recurrent Network for Deep Image Deblurring (CVPR 2018 · 1,365 citations)
- Efficient Multi-scale Network with Learnable Discrete Wavelet Transform for Blind Motion Deblurring (CVPR 2024 · 78 citations)
- Unmixing Diffusion for Self-Supervised Hyperspectral Image Denoising (CVPR 2024 · 38 citations)
Learned image compression · 139 papers
Keywords: compression, distortion, coding
Approaches in this cluster:
- Learned image compression (104 papers)
Compress images with learned hyperprior entropy models and lossless or lossy coding. - Perceptual diffusion image compression (35 papers)
Use diffusion decoders to improve perceptual quality at ultra-low bitrates.
Most cited and most cited since 2024:
- Variational image compression with a scale hyperprior (ICLR 2018 · 1,378 citations)
- End-to-end Optimized Image Compression (ICLR 2017 · 1,342 citations)
- Generative Latent Coding for Ultra-Low Bitrate Image Compression (CVPR 2024 · 27 citations)
- Learned Image Compression with Dictionary-based Entropy Model (CVPR 2025 · 23 citations)
Related topics in Image generation
- Autoregressive and latent generation (2,100)
- GANs (780)
- Super-resolution (771)
- Editing and style transfer (1,430)
- Fast diffusion sampling (2,155)
- Text-to-image diffusion (1,916)
