Research map / Image generation / Image restoration
Beyond Ground-Truth: Leveraging Image Quality Priors for Real-World Image Restoration
Fengyang Xiao, Peng Hu, Lei Xu, XingE Guo, Guanyi Qin, Yuqi Shen, Chengyu Fang, Rihan Zhang, Chunming He, Sina Farsiu
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 NR-IQA score as condition to the code-prediction Transformer (ablation) | codebook restoration · WebPhoto-Test | TOPIQ-G: 0.757 -> 0.790 | Improved |
| add dual codebook (common + HQ+) (ablation) | score-conditioned model · WebPhoto-Test | TOPIQ-G: 0.790 -> 0.806 | Improved |
| add NR-IQA quality optimization loss (ablation) | score-conditioned dual-codebook model · WebPhoto-Test | TOPIQ-G: 0.806 -> 0.822 | Improved |
| add dual codebook (ablation) | score-conditioned model · WebPhoto-Test | Musiq-K: 75.95 -> 76.04 | No clear effect |
| add all three components (score condition, dual codebook, quality optimization) (ablation) | codebook restoration · WebPhoto-Test | Q-Align: 3.78 -> 4.23 | Improved |
| add all components, inference cost (ablation) | codebook restoration · WebPhoto-Test | Inference time (ms): 5.27 -> 5.63 | Hurt |
| optimize NR-IQA score in discrete codebook space instead of continuous space (ablation) | quality optimization · Fig. 7 visual study | human-perceived realism | Improved |
| plug quality prior into WaveFace | WaveFace · LFW-Test | TOPIQ-G: 0.786 -> 0.802 | Improved |
| plug quality prior into Interlcm | Interlcm · LFW-Test | TOPIQ-G: 0.831 -> 0.850 | Improved |
| plug quality prior into Reti-Diff | Reti-Diff · LOL-v1 | FID: 49.14 -> 42.15 | Improved |
| plug quality prior into CIDNet | CIDNet · LOL-v2-real | FID: 48.04 -> 43.33 | Improved |
| plug quality prior into CIDNet for detection | CIDNet · ExDark | mean AP: 73.6 -> 75.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Reti-Diff — Reti-Diff: Illumination Degradation Image Restoration with Retinex-based Latent Diffusion Model (ICLR 2025) · better on 5 of 5 shared results
- Interlcm — $InterLCM$: Low-Quality Images as Intermediate States of Latent Consistency Models for Effective Blind Face Restoration (ICLR 2025) · better on 3 of 3 shared results
- WaveFace — WaveFace: Authentic Face Restoration with Efficient Frequency Recovery (CVPR 2024) · better on 2 of 2 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
- UARE: A Unified Vision-Language Model for Image Quality Assessment, Restoration, and Enhancement (CVPR 2026 · similarity 0.91)
- Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild (CVPR 2024 · similarity 0.89)
- Augmenting Perceptual Super-Resolution via Image Quality Predictors (CVPR 2025 · similarity 0.89)
- IRPO: Boosting Image Restoration via Post-training GRPO (NeurIPS 2026 · similarity 0.88)
- Posterior-Mean Rectified Flow: Towards Minimum MSE Photo-Realistic Image Restoration (ICLR 2025 · similarity 0.88)
- Continuous Expert Assembly: Instance-Conditioned Low-Rank Residuals for All-in-One Image Restoration (NeurIPS 2026 · 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.
