Research map / Image generation / Super-resolution
Towards Bidirectional Arbitrary Image Rescaling: Joint Optimization and Cycle Idempotence
Zhihong Pan, Baopu Li, Dongliang He, Mingde Yao, Wenhao Wu, Tianwei Lin, Xin Li, Errui Ding
Check 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 |
|---|---|---|---|
| jointly optimize arbitrary downscaling and upscaling as one bidirectional rescaling model | BAIRNet · arbitrary scales ×1.5 and ×2.5 on Set5, Urban100, and DIV2K | PSNR: 40.11 vs 39.78; 36.62 vs 33.99; 37.46 vs 36.60 at ×2.5 | Improved |
| use BAIRNet for ×3.5 rescaling | BAIRNet† · ×3.5, Set5 | PSNR: 36.85 vs 37.12 | Hurt |
| use BAIRNet for ×3.5 rescaling | BAIRNet† · ×3.5, Urban100 and DIV2K | PSNR: 32.71 vs 31.84; 33.87 vs 33.84 | Improved |
| train with downscaling subpixel weight function (D-SWF) (ablation) | BAIRNet · DIV2K, ×4 | PSNR: 31.19 -> 31.42 | Improved |
| weaken LR reference supervision from pixel-level L₂ to Lᶜ₂ or mean-color Lᵐ₂ (ablation) | BAIRNet · DIV2K, ×4 | PSNR: 31.78 -> 32.12; 31.78 -> 32.13 | Improved |
| remove LR reference supervision (ablation) | BAIRNet · DIV2K, ×4 | PSNR: 31.78 -> 32.13 | Improved |
| use Lᶜ₂ supervision on Cb and Cr channels to suppress false-color artifacts (ablation) | BAIRNet · DIV2K, ×4 | PSNR: 32.13 -> 32.12 | Hurt |
| remove scale normalization from the reconstruction loss (ablation) | BAIRNet · DIV2K, ×2 | PSNR: 40.15 -> 39.93 | Hurt |
| replace scale-normalized Lₛ reconstruction loss with L₁ (ablation) | BAIRNet · DIV2K, ×4 | PSNR: 32.13 -> 32.14 | No clear effect |
| fine-tune using losses over N=3 repeated rescaling cycles (ablation) | BAIRNet · DIV2K, ×4, cycles 2–5 | PSNR improvement over base BAIRNet: 0.16; 0.27; 0.40; 0.54 | Improved |
| fine-tune using losses over N=5 repeated rescaling cycles (ablation) | BAIRNet · DIV2K, ×4, cycle 1 versus cycles 2–5 | PSNR improvement over base BAIRNet: 0.05; 0.15; 0.28; 0.44; 0.61 | Mixed |
| apply asymmetric-scale fine-tuning (ablation) | BAIRNet · asymmetric scales, five benchmark datasets | PSNR: 36.00 -> 36.66 on Set14, ×2/×3 | Improved |
Closest papers by meaning
- Task-Aware Image Downscaling (ECCV 2018 · similarity 0.89)
- Deep Generative Model based Rate-Distortion for Image Downscaling Assessment (CVPR 2024 · similarity 0.88)
- Arbitrary-Scale Image Generation and Upsampling using Latent Diffusion Model and Implicit Neural Decoder (CVPR 2024 · similarity 0.87)
- Block-Based Multi-Scale Image Rescaling (AAAI 2025 · similarity 0.86)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.86)
- Navigating Beyond Dropout: An Intriguing Solution towards Generalizable Image Super Resolution (CVPR 2024 · similarity 0.86)
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.
