Research map / Image generation / Image restoration
Learning Steerable Function for Efficient Image Resampling
Jiacheng Li, Chang Chen, Wei Huang, Zhiqiang Lang, Fenglong Song, Youliang Yan, Zhiwei Xiong
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 |
|---|---|---|---|
| Use locally adapted steerable resampling functions | LeRF · Set5, ×2 upsampling | PSNR: 35.71 vs 30.75 | Improved |
| Replace the steerable Gaussian with learned isotropic Keys cubic (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 35.71 vs 34.68 | Hurt |
| Train on ×2 rather than ×4 downsampled pairs (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 35.71 vs 35.67 | No clear effect |
| Train on mixed-scale data pairs (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 35.71 vs 35.67 | No clear effect |
| Remove LUT acceleration and use the learned network (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 35.71 vs 36.11 | Improved |
| Add LUT acceleration (ablation) | LeRF · Set5, ×2 upsampling | MACs: 21.75G vs 57.94M | Improved |
| Add directional ensemble instead of rotation ensemble (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 34.44 vs 35.71 | Improved |
| Add edge-sensitive “C” and “X” indexing patterns (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 34.90 vs 35.71 | Improved |
| Add the pre-filtering stage (ablation) | LeRF · Set5, ×2 upsampling | PSNR: 34.43 vs 35.71 | Improved |
| Add adaptive resampling functions after pre-filtering (ablation) | pre-filtering-only variant · Set5, ×2 upsampling | PSNR: 34.69 vs 35.71 | Improved |
| Use LeRF instead of bicubic interpolation | Set14 ×4 upsampling · Set14, ×4 upsampling | running time: 126 -> 110 | Improved |
| Use LeRF instead of LIIF | Set14 ×4 upsampling · Set14, ×4 upsampling | PSNR: 28.65 vs 27.35 | Hurt |
Closest papers by meaning
- LeRF: Learning Resampling Function for Adaptive and Efficient Image Interpolation (TPAMI 2025 · similarity 0.96)
- CUF: Continuous Upsampling Filters (CVPR 2023 · similarity 0.84)
- Learning Affinity-Aware Upsampling for Deep Image Matting (CVPR 2021 · similarity 0.83)
- Low Distortion Block-Resampling with Spatially Stochastic Networks (NeurIPS 2020 · similarity 0.83)
- Enhancing Video Super-Resolution via Implicit Resampling-based Alignment (CVPR 2024 · similarity 0.81)
- Plug-and-Play Tri-Branch Invertible Block for Image Rescaling (AAAI 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.
