Research map / Image generation / Super-resolution
Learning Degradation Representations for Image Deblurring
Dasong Li, Yi Zhang, Ka Chun Cheung, Xiaogang Wang, Hongwei Qin, Hongsheng Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add the learned degradation encoder and representations (ablation) | Ours (deblurring generator) · GoPro test | PSNR: 32.81 -> 33.28 | Improved |
| add joint reblurring generator for learning degradation representations (ablation) | Ours · GoPro test | PSNR: 33.09 -> 33.28 | Improved |
| add blur-aware loss (ablation) | Ours · GoPro test | PSNR: 33.21 -> 33.28 | Improved |
| inject degradation at multiple scales instead of only the lowest resolution (ablation) | Ours · GoPro test | PSNR: 32.80 -> 33.28 | Improved |
| use spatially adaptive modulation instead of concatenation at skip connections (ablation) | Ours · GoPro test | PSNR: 33.05 -> 33.28 | Improved |
| inject via skip-connection modulation instead of concatenating degradation map at the network input (ablation) | Ours · GoPro test | PSNR: 32.98 -> 33.28 | Improved |
| use learned degradation representations vs MPRNet-patch256 on blurriest 10% images | Ours · GoPro test (blurriest 10%) | PSNR: 29.31 -> 29.65 | Improved |
| use MSDI-Net with degradation representations vs MPRNet-patch256 compute | Ours · GoPro test | MACs (G): 760.11 -> 336.43 | Improved |
| apply degradation representations trained on GoPro directly to RealBlur | Ours · RealBlur test | PSNR: 32.12 -> 32.35 | Improved |
| use Ours instead of HINet | HINet · GoPro test | PSNR: 32.71 -> 33.28 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HINet — HiNet: Deep Image Hiding by Invertible Network (ICCV 2021) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Unsupervised Degradation Representation Aware Transform for Real-World Blind Image Super-Resolution (AAAI 2025 · similarity 0.90)
- Deblurring Using Analysis-Synthesis Networks Pair (CVPR 2020 · similarity 0.89)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.89)
- Unsupervised Degradation Representation Learning for Blind Super-Resolution (CVPR 2021 · similarity 0.89)
- DDR: Exploiting Deep Degradation Response as Flexible Image Descriptor (NeurIPS 2024 · similarity 0.88)
- Visual-Instructed Degradation Diffusion for All-in-One Image Restoration (CVPR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
