Research map / Image generation / Image restoration
LeRF: Learning Resampling Function for Adaptive and Efficient Image Interpolation
Jiacheng Li, Chang Chen, 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 learned anisotropic adaptive Gaussian instead of fixed Gaussian (ablation) | LeRF-G · Set5 x2 | PSNR: 30.75 -> 35.71 | Improved |
| use adaptive anisotropic Gaussian instead of isotropic Keys cubic (ablation) | LeRF-G · Set5 x2 | PSNR: 34.68 -> 35.71 | Improved |
| add LUT acceleration to LeRF-G (ablation) | LeRF-G · Set5 x2 | PSNR: 36.11 -> 35.71 | Mixed |
| add DE (directional encoding) instead of RE (ablation) | LeRF-G · Set5 x2 | PSNR: 34.44 -> 35.71 | Improved |
| add CX pattern (ablation) | LeRF-G · Set5 x2 | PSNR: 34.90 -> 35.71 | Improved |
| add pre-processing stage (ablation) | LeRF-G · Set5 x2 | PSNR: 34.43 -> 35.71 | Improved |
| train on x2 data instead of x4 data (ablation) | LeRF-G · Set5 x4 | PSNR: 30.15 -> 29.91 | Hurt |
| train on mixed-scale data instead of x4 data (ablation) | LeRF-G · Set5 x3 | PSNR: 32.02 -> 32.26 | No clear effect |
| replace LUT multi-branch DNN with convolution-only LeRF-Net (ablation) | LeRF-Net · Set5 x2 | PSNR: 36.11 -> 36.23 | Improved |
| increase channels from 16 to 64 (ablation) | LeRF-Net · Set5 x2 | PSNR: 36.44 -> 36.57 | Improved |
| increase channels from 64 to 96 (ablation) | LeRF-Net · Set5 x2 | PSNR: 36.57 -> 36.52 | No clear effect |
| combine LeRF-G with SR-LUT | SR-LUT · Set5 x2 | PSNR: 35.53 -> 35.53 | No clear effect |
Closest papers by meaning
- Learning Steerable Function for Efficient Image Resampling (CVPR 2023 · similarity 0.96)
- Learning Affinity-Aware Upsampling for Deep Image Matting (CVPR 2021 · similarity 0.85)
- CUF: Continuous Upsampling Filters (CVPR 2023 · similarity 0.84)
- Linearized Multi-Sampling for Differentiable Image Transformation (ICCV 2019 · similarity 0.84)
- Aug-NeRF: Training Stronger Neural Radiance Fields With Triple-Level Physically-Grounded Augmentations (CVPR 2022 · similarity 0.84)
- Deep Network Interpolation for Continuous Imagery Effect Transition (CVPR 2019 · 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.
