Research map / Image generation / Super-resolution
Neural Discrimination-Prompted Transformers for Efficient UHD Image Restoration and Enhancement
Cong Wang, Jinshan Pan, Liyan Wang, Wei William Wang, Yang Yang
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 NDP in NDPA and NDPN (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 26.811 -> 27.159 | Improved |
| add NDP in NDPA (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 26.183 -> 27.159 | Improved |
| add NDP in NDPN (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 27.026 -> 27.159 | Improved |
| use refined NDP instead of direct stride-conv features (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 26.138 -> 27.159 | Improved |
| insert NDP inside attention and FFN rather than before each block (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 26.161 -> 27.159 | Improved |
| use SR-guided reconstruction instead of cascaded reconstruction (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 26.375 -> 27.159 | Improved |
| use SR-guided reconstruction instead of cascaded reconstruction (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | SSIM: 0.9043 -> 0.9285 | Improved |
| add SR loss (second loss term) (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | PSNR: 27.125 -> 27.159 | Improved |
| increase channels beyond 16 (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | SSIM | No clear effect |
| shuffle down factor 4 instead of 8 (ablation) | UHDPromer · UHD-LL (trained on UHD-LL) | SSIM | No clear effect |
| apply UHDPromer to general-size deblurring | UHDPromer · GoPro (general size) | PSNR: 29.487 vs FFTformer 35.798 | Hurt |
| apply UHDPromer to general-size dehazing | UHDPromer · SOTS-Indoor (general size) | PSNR: 32.4778 vs UHDformer 35.7413 | Hurt |
Closest papers by meaning
- Correlation Matching Transformation Transformers for UHD Image Restoration (AAAI 2024 · similarity 0.88)
- UHD-processer: Unified UHD Image Restoration with Progressive Frequency Learning and Degradation-aware Prompts (CVPR 2025 · similarity 0.88)
- From Zero to Detail: Deconstructing Ultra-High-Definition Image Restoration from Progressive Spectral Perspective (CVPR 2025 · similarity 0.86)
- UP-Restorer: When Unrolling Meets Prompts for Unified Image Restoration (AAAI 2025 · similarity 0.85)
- DreamUHD: Frequency Enhanced Variational Autoencoder for Ultra-High-Definition Image Restoration (AAAI 2025 · similarity 0.85)
- Dual Prompting Image Restoration with Diffusion Transformers (CVPR 2025 · similarity 0.85)
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.
