Research map / Image generation / Super-resolution
Toward Real-World Single Image Super-Resolution: A New Benchmark and a New Model
Jianrui Cai, Hui Zeng, Hongwei Yong, Zisheng Cao, 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 |
|---|---|---|---|
| train on RealSR dataset instead of bicubic degradation (BD) data | RCAN · RealSR test set (x2) | PSNR (dB): 32.91 -> 33.87 | Improved |
| train on RealSR dataset instead of BD data | RCAN · RealSR test set (x4) | PSNR (dB): 28.28 -> 28.88 | Improved |
| train on RealSR dataset instead of BD data | VDSR · RealSR test set (x4) | PSNR (dB): 28.03 -> 28.63 | Improved |
| train on multiple simulated degradations (MD) instead of BD | SRResNet · RealSR test set (x4) | PSNR (dB): 28.09 -> 28.12 | No clear effect |
| train on BD data instead of using bicubic interpolation | VDSR · RealSR test set (x4) | PSNR (dB): 27.99 -> 28.03 | No clear effect |
| use kernel prediction (k=5) instead of direct pixel synthesis (ablation) | DPS · RealSR test set (x4) | PSNR (dB): 28.69 -> 28.74 | Improved |
| increase KPN kernel size from 5 to 19 (ablation) | KPN · RealSR test set (x4) | PSNR (dB): 28.74 -> 28.90 | Improved |
| add Laplacian pyramid decomposition to KPN with k=5 (ablation) | KPN, k=5 · RealSR test set (x4) | PSNR (dB): 28.74 -> 28.92 | Improved |
| add Laplacian pyramid to KPN, with lower cost than k=19 (ablation) | KPN, k=19 · RealSR cost (1200x2200 image) | running time (s): 0.2748 -> 0.1813 | Improved |
| use LP-KPN instead of RCAN | RCAN · RealSR test set (x4) | parameters (M): 32.71M -> 5.731M | Improved |
| train LP-KPN on Nikon and test on Canon instead of in-camera | LP-KPN · Cross-camera, tested on Canon (x4) | PSNR (dB): 29.48 -> 29.10 | Hurt |
| train RCAN on other camera instead of same camera | RCAN · Cross-camera, tested on Canon (x4) | PSNR (dB): 29.46 -> 29.04 | Hurt |
Closest papers by meaning
- Learning Controllable Degradation for Real-World Super-Resolution via Constrained Flows (ICML 2023 · similarity 0.92)
- Kernel Modeling Super-Resolution on Real Low-Resolution Images (ICCV 2019 · similarity 0.92)
- Learning a Single Convolutional Super-Resolution Network for Multiple Degradations (CVPR 2018 · similarity 0.92)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.90)
- Towards Realistic Data Generation for Real-World Super-Resolution (ICLR 2025 · similarity 0.90)
- Unsupervised Diffusion-Based Degradation Modeling for Real-World Super-Resolution (AAAI 2025 · similarity 0.89)
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.
