Research map / Image generation / Super-resolution
Component Divide-and-Conquer for Real-World Image Super-Resolution
Pengxu Wei, Ziwei Xie, Hannan Lu, Zongyuan Zhan, Qixiang Ye, Wangmeng Zuo, Liang Lin
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 intermediate supervision to HGSR (ablation) | HGSR · DRealSR x4 | PSNR (dB): 31.95 -> 32.13 | Improved |
| replace conv layers between HG modules with RIBs (ablation) | HGSR · DRealSR x4 | PSNR (dB): 32.13 -> 32.15 | No clear effect |
| add component-attentive blocks (CAB) (ablation) | HGSR+RIB · DRealSR x4 | PSNR (dB): 32.15 -> 32.27 | Improved |
| add gradient-weighted loss (ablation) | HGSR+RIB+CAB · DRealSR x4 | PSNR (dB): 32.27 -> 32.42 | Improved |
| add the corner branch to flat+edge guidance (ablation) | CDC · DRealSR x4 | PSNR (dB): 32.23 -> 32.42 | Improved |
| use flat, edge and corner guidance instead of edge+corner (ablation) | CDC · DRealSR x4 | PSNR (dB): 32.40 -> 32.42 | No clear effect |
| replace L1 with gradient-weighted loss (ablation) | EDSR · DRealSR x4 | PSNR (dB): 32.03 -> 32.27 | Improved |
| replace L1 with gradient-weighted loss (ablation) | SRResNet · DRealSR x4 | PSNR (dB): 31.63 -> 31.93 | Improved |
| replace L1 with gradient-weighted loss (ablation) | HGSR · DRealSR x4 | PSNR (dB): 32.15 -> 32.25 | Improved |
| increase HG blocks from 2 to 6 (ablation) | HGSR · DRealSR x4 | PSNR (dB): 31.55 -> 31.95 | Improved |
| increase HG blocks from 6 to 8 (ablation) | HGSR · DRealSR x4 | PSNR (dB): 31.95 -> 31.94 | No clear effect |
| train on real-degraded data instead of bicubic data when testing on real LR | CDC · DRealSR x4 (test on real LR) | PSNR (dB): 30.63 -> 32.42 | Improved |
Closest papers by meaning
- Towards Realistic Data Generation for Real-World Super-Resolution (ICLR 2025 · similarity 0.88)
- DORNet: A Degradation Oriented and Regularized Network for Blind Depth Super-Resolution (CVPR 2025 · similarity 0.88)
- SEAL: A Framework for Systematic Evaluation of Real-World Super-Resolution (ICLR 2024 · similarity 0.87)
- DegBins: Degradation-Driven Binning for Depth Super-Resolution (NeurIPS 2026 · similarity 0.87)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.87)
- StructSR: Refuse Spurious Details in Real-World Image Super-Resolution (AAAI 2025 · similarity 0.87)
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.
