Research map / Image generation / Super-resolution
Zero-Shot Inference-Time Rectification for Real-World Arbitrary-Scale Super-Resolution
Yifan Zuo, tianlin zhu, zhenlong xia, Jiebin Yan, Xiaoshui Huang, Sanqian Li, Yuming Fang, Qiang Wu
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 CDR inference proxy to LIIF | LIIF · RealSR x4 | PSNR: +0.8238 | Improved |
| add CDR inference proxy to CiaoSR | CiaoSR · RealSR x4 | LPIPS: -0.1852 | Improved |
| add CDR inference proxy to LIIF | LIIF · DRealSR x4 | SSIM: -0.0296 | Hurt |
| add CDR inference proxy to LIIF | LIIF · COZ x4 | PSNR: +3.89 dB | Improved |
| add CDR inference proxy to CiaoSR | CiaoSR · COZ x2 | SSIM: +0.0684 | Improved |
| use continuous-scale conditioning instead of fixed x4 training (ablation) | CDR with LIIF · COZ x2 | PSNR: 28.2358 -> 31.7725 | Improved |
| use continuous-scale conditioning instead of fixed x4 training (ablation) | CDR with LIIF · COZ x6 | PSNR: 22.9156 -> 26.4225 | Improved |
| train CDR on fixed x2 scale instead of no CDR (ablation) | CDR with LIIF · COZ x2 | PSNR: 27.8485 -> 28.2575 | No clear effect |
| use channel-wise modulation on Kdc instead of feature addition (ablation) | CDR with LIIF · COZ x2 | PSNR: 28.0059 -> 31.7725 | Improved |
| use channel-wise instead of global modulation on Kdc (ablation) | CDR with LIIF · COZ x2 | PSNR: 28.3928 -> 31.7725 | Improved |
| inject Fsca into kernel generation (Kdc) instead of Kd, channel-wise (ablation) | CDR with LIIF · COZ x2 | PSNR: 28.2760 -> 31.7725 | Improved |
| use feature concatenation instead of addition (ablation) | CDR with LIIF · COZ x3 | PSNR: 26.3641 -> 26.8999 | Improved |
Closest papers by meaning
- Learning Heterogeneous Degradation Representation for Real-World Super-Resolution (ICLR 2026 · similarity 0.89)
- NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution (NeurIPS 2026 · similarity 0.88)
- Unsupervised Diffusion-Based Degradation Modeling for Real-World Super-Resolution (AAAI 2025 · similarity 0.88)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.88)
- Towards Realistic Data Generation for Real-World Super-Resolution (ICLR 2025 · similarity 0.88)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · 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.
