Research map / Image generation / Super-resolution
From Face to Natural Image: Learning Real Degradation for Blind Image Super-Resolution
Xiaoming Li, Chaofeng Chen, Xianhui Lin, Wangmeng Zuo, Lei Zhang
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 disentanglement loss to ReDegNet training (ablation) | F2N-ESRGAN · RealLQSet | NIQE: 6.27 -> 4.93 | Improved |
| add the disentanglement loss to ReDegNet training (ablation) | F2N-ESRGAN · RealSR-Canon | LPIPS: .463 -> .362 | Improved |
| add the disentanglement loss to ReDegNet training (ablation) | F2N-ESRGAN · RealSR-Canon | PSNR: 24.63 -> 25.57 | Improved |
| use paired LQ/HQ face images (supervised) instead of unpaired data (ablation) | F2N-ESRGAN · RealLQSet | NIQE: 5.93 -> 4.93 | Improved |
| use paired LQ/HQ face images (supervised) instead of unpaired data (ablation) | F2N-ESRGAN · DRealSR | LPIPS: .404 -> .373 | Improved |
| use paired LQ/HQ face images (supervised) instead of unpaired data (ablation) | F2N-ESRGAN · RealSRSet | NIQE: 5.81 -> 4.85 | Improved |
| learned face-derived degradation (F2N-ESRGAN) versus handcrafted degradation (Real-ESRGAN) | ESRGAN · RealLQSet | NIQE: 5.24 -> 4.93 | Improved |
| learned face-derived degradation (F2N-ESRGAN) versus RealSR degradation | ESRGAN · RealSRSet | NIQE: 4.82 -> 4.85 | Hurt |
| learned face-derived degradation (F2N-ESRGAN) versus BSRGAN | ESRGAN · RealSR-Canon | PSNR: 25.61 -> 25.57 | No clear effect |
| learned face-derived degradation (F2N-ESRGAN) versus BSRGAN | ESRGAN · DRealSR | LPIPS: .380 -> .373 | Improved |
Closest papers by meaning
- Continuous Degradation Modeling via Latent Flow Matching for Real-World Super-Resolution (AAAI 2026 · similarity 0.90)
- Unsupervised Diffusion-Based Degradation Modeling for Real-World Super-Resolution (AAAI 2025 · similarity 0.89)
- Unsupervised Degradation Representation Aware Transform for Real-World Blind Image Super-Resolution (AAAI 2025 · similarity 0.89)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.88)
- DSL-FIQA: Assessing Facial Image Quality via Dual-Set Degradation Learning and Landmark-Guided Transformer (CVPR 2024 · similarity 0.88)
- Low-Light Face Super-resolution via Illumination, Structure, and Texture Associated Representation (AAAI 2024 · 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.
