Research map / Image generation / Super-resolution
Better "CMOS" Produces Clearer Images: Learning Space-Variant Blur Estimation for Blind Image Super-Resolution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use CMOS for blind SR instead of MANet | NYUv2-BSR · NYUv2-BSR test (5-group average) | PSNR/SSIM: 30.12/0.9106 -> 32.03/0.9154 | Helped |
| Use CMOS for blind SR instead of MANet | Cityscapes-BSR · Cityscapes-BSR test (5-group average) | PSNR/SSIM: 34.25/0.9287 -> 35.58/0.9383 | Helped |
| Train with space-variant blur rather than space-invariant blur (ablation) | MANet · NYUv2-BSR | PSNR/SSIM: +10.62 dB; +0.2266 | Helped |
| Add flow-based upsampling to GIA (ablation) | CMOS · NYUv2-BSR | PSNR: 23.21 -> 23.42 | Helped |
| Add channel interaction to flow-based upsampling (ablation) | CMOS · NYUv2-BSR | PSNR: 23.42 -> 24.24 | Helped |
| Add spatial interaction to flow-based upsampling and channel interaction (ablation) | CMOS · NYUv2-BSR | PSNR: 24.24 -> 24.52 | Helped |
| Add spatial interaction to flow-based upsampling and channel interaction (ablation) | CMOS · NYUv2-BSR | mIoU: 36.25 -> 35.61 | Hurt |
| Add semantic maps to the SR model input (ablation) | RRDB-SFT with ground-truth degradation maps · NYUv2-BSR | PSNR/SSIM: +0.34 dB; +0.0022 | Helped |
| Jointly estimate blur and semantic maps instead of predicting them separately (ablation) | CMOS · NYUv2-BSR | mIoU: +1.66 | Helped |
| Jointly estimate blur and semantic maps instead of predicting them separately (ablation) | CMOS with RRDB-SFT · NYUv2-BSR | PSNR/SSIM: +1.28/+0.002 | Helped |
| Jointly estimate blur and semantic maps instead of predicting them separately (ablation) | CMOS · NYUv2-BSR | blur-map PSNR/SSIM: 24.58/0.8393 -> 24.52/0.8340 | Hurt |
| Add auxiliary supervision to the multi-scale structure (ablation) | CMOS · NYUv2-BSR | PSNR: 24.14 -> 24.52; 0.38 | Helped |
Closest papers by meaning
- Adaptive Dropout: Unleashing Dropout across Layers for Generalizable Image Super-Resolution (CVPR 2025 · similarity 0.84)
- Navigating Beyond Dropout: An Intriguing Solution towards Generalizable Image Super Resolution (CVPR 2024 · similarity 0.83)
- AutoLUT: LUT-Based Image Super-Resolution with Automatic Sampling and Adaptive Residual Learning (CVPR 2025 · similarity 0.81)
- Augmenting Perceptual Super-Resolution via Image Quality Predictors (CVPR 2025 · similarity 0.81)
- Zero-Shot Blind-spot Image Denoising via Implicit Neural Sampling (CVPR 2025 · similarity 0.81)
- A Physics-Informed Blur Learning Framework for Imaging Systems (CVPR 2025 · similarity 0.81)
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.