Research map / Image generation / Super-resolution
SuperF: Neural Implicit Fields for Multi-Image Super-Resolution
Sander Riisøen Jyhne, Christian Igel, Morten Goodwin, Per-Arne Andersen, Serge Belongie, Nico Lang
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 |
|---|---|---|---|
| optimize sub-pixel alignment jointly with the shared INR (vs multi-frame without alignment) (ablation) | SuperF · SatSynthBurst x4 | PSNR (dB): 28.11 -> 32.94 | Improved |
| use multiple frames without alignment instead of a single frame (ablation) | SuperF · SatSynthBurst x4 | PSNR (dB): 30.42 -> 28.11 | Hurt |
| add Fourier feature positional encoding (ablation) | SuperF INR · SatSynthBurst x4 | PSNR (dB): 20.33 -> 30.42 | Improved |
| directly parameterize affine transforms instead of an MLP (ablation) | NIR · SatSynthBurst x4 | PSNR (dB): 24.63 -> 26.14 | Improved |
| add supersampling alone to NIR (ablation) | NIR · SatSynthBurst x4 | PSNR (dB): 24.63 -> 24.76 | No clear effect |
| combine direct T and supersampling and fixed base frame (ablation) | NIR · SatSynthBurst x4 | PSNR (dB): 24.63 -> 32.94 | Improved |
| add fixed base frame without direct parameterization (with supersampling) (ablation) | NIR · SatSynthBurst x4 | alignment error: 0.079 -> 1.324 | Hurt |
| use GNLL instead of MSE loss on satellite bursts (ablation) | SuperF · SatSynthBurst x4 | PSNR (dB): 32.94 -> 34.03 | Improved |
| use GNLL instead of MSE on ground-level bursts (ablation) | SuperF · SyntheticBurst x4 | PSNR (dB): 27.90 -> 27.47 | No clear effect |
| use GNLL instead of MSE on noisy cloudy series (ablation) | SuperF · WorldStrat-bitter | PSNR (dB): 27.42 -> 28.46 | Improved |
| use GNLL instead of MSE on clean series (ablation) | SuperF · WorldStrat-sweet | PSNR (dB): 27.86 -> 27.88 | No clear effect |
| increase max frame shift from 1 to 4 LR pixels at x8 (ablation) | SuperF · SatSynthBurst x8 | PSNR (dB): 29.49 -> 18.28 | Hurt |
Closest papers by meaning
- NeurOp-Diff: Continuous Remote Sensing Image Super-Resolution via Neural Operator Diffusion (ICCV 2025 · similarity 0.89)
- Noise-Free Optimization in Early Training Steps for Image Super-resolution (AAAI 2024 · similarity 0.87)
- Deep Burst Super-Resolution (CVPR 2021 · similarity 0.87)
- Hitchhiker's Guide to Super-Resolution: Introduction and Recent Advances (TPAMI 2023 · similarity 0.87)
- HIIF: Hierarchical Encoding based Implicit Image Function for Continuous Super-resolution (CVPR 2025 · similarity 0.86)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (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.
