Research map / Image generation / Super-resolution
Super-Resolution Neural Operator
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 |
|---|---|---|---|
| Replace the LTE decoder with SRNO's Galerkin-type attention decoder | EDSR-baseline continuous SR · DIV2K validation, ×2 through ×30 | PSNR (dB): 34.72 -> 34.85; 31.02 -> 31.11; 29.04 -> 29.16; 26.81 -> 26.90; 23.78 -> 23.84; 22.23 -> 22.29; 21.24 -> 21.27; 20.53 -> | Helped |
| Replace the LTE decoder with SRNO's Galerkin-type attention decoder | RDN continuous SR · Set5, Set14, B100, and Urban100, ×4 and ×8 | PSNR (dB): 32.61 -> 32.69; 27.26 -> 27.28; 28.88 -> 28.97; 25.16 -> 25.26; 27.77 -> 27.83; 24.95 -> 24.99; 26.81 -> 26.98; 22.88 -> | Helped |
| Use nearest-neighborhood interpolation and position embedding instead of SRNO's interpolation-free lifting (ablation) | SRNO · Set14 ×2/×3/×4/×6/×8 | PSNR (dB): 33.83 -> 33.81; 30.50 -> 30.47; 28.79 -> 28.76; 26.55 -> 26.55; 25.05 -> 25.00 | Hurt |
| Reduce attention heads from 16 to 8 (ablation) | SRNO · Set14 ×2/×3/×4/×6/×8 | PSNR (dB): 33.83 -> 33.87; 30.50 -> 30.47; 28.79 -> 28.76; 26.55 -> 26.56; 25.05 -> 25.02 | Mixed |
| Reduce latent dimension from 256 to 128 (ablation) | SRNO · Set14 ×2/×3/×4/×6/×8 | PSNR (dB): 33.83 -> 33.80; 30.50 -> 30.45; 28.79 -> 28.75; 26.55 -> 26.53; 25.05 -> 25.00 | Hurt |
| Reduce kernel-integral iterations from 2 to 1 (ablation) | SRNO · Set14 ×2/×3/×4/×6/×8 | PSNR (dB): 33.83 -> 33.82; 30.50 -> 30.45; 28.79 -> 28.75; 26.55 -> 26.55; 25.05 -> 25.00 | Hurt |
| Reduce kernel-integral iterations from 2 to 1 (ablation) | SRNO · Set14 ×2/×3/×4/×6/×8 | Time (ms): 149 -> 117 | Helped |
| Use sequential rather than random coordinate sampling during training (ablation) | SRNO · DIV2K validation, ×2 through ×30 | PSNR (dB): 34.85 -> 34.79; 31.11 -> 31.07; 29.16 -> 29.09; 26.90 -> 26.84; 23.84 -> 23.80; 22.29 -> 22.26; 21.27 -> 21.24; 20.56 -> | Hurt |
| Increase the number of sampled HR pixels (ablation) | SRNO · DIV2K training; several SR scales | PSNR | Helped |
Closest papers by meaning
- Hierarchical Neural Operator Transformer with Learnable Frequency-aware Loss Prior for Arbitrary-scale Super-resolution (ICML 2024 · similarity 0.90)
- DiffFNO: Diffusion Fourier Neural Operator (CVPR 2025 · similarity 0.86)
- HIIF: Hierarchical Encoding based Implicit Image Function for Continuous Super-resolution (CVPR 2025 · similarity 0.86)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.85)
- MgNO: Efficient Parameterization of Linear Operators via Multigrid (ICLR 2024 · similarity 0.85)
- Implicit Representations via Operator Learning (ICML 2024 · similarity 0.84)
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.