Research map / Image generation / Image restoration
Learning Hierarchical Line Buffer for Image Processing
Jiacheng Li, Feiran Li, Daisuke Iso
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 hierarchical line buffers instead of naive LSTM (ablation) | line-buffer network · SID RAW denoising x100 | PSNR: 39.69 -> 41.13 | Improved |
| add more LSTM modules (ablation) | LineLSTM · SID RAW denoising x100 | PSNR: 39.69 -> 39.94 | No clear effect |
| add progressive enhancement with channel split (ablation) | LineLSTM · SID RAW denoising x100 | PSNR: 39.69 -> 40.87 | Improved |
| add strip-shaped convolutions (ablation) | progressive-enhance LineLSTM · SID RAW denoising x100 | PSNR: 40.87 -> 40.98 | Improved |
| use strip convolutions instead of ordinary convolutions (ablation) | full method · SID RAW denoising x100 | PSNR: 41.53 -> 41.69 | Improved |
| add more intra-line feature extractor blocks (ablation) | full method · SID RAW denoising x100 | PSNR: 41.69 -> 41.91 | Improved |
| use fewer intra-line blocks (ablation) | full method · SID RAW denoising x100 | PSNR: 41.69 -> 41.58 | Hurt |
| add the full hierarchy levels instead of fewer levels (ablation) | full method · SID RAW denoising x100 | PSNR: 40.98 -> 41.69 | Improved |
| add inter-line design to the baseline (ablation) | Baseline · SID RAW denoising x100 | PSNR: 40.12 -> 41.69 | Improved |
| increase lines per group from 4 to 16 (ablation) | Baseline · SID RAW denoising x100 | PSNR: 38.78 -> 39.72 | Improved |
| increase lines per group from 4 to 16 (ablation) | Ours · SID RAW denoising x100 | PSNR: 41.68 -> 41.89 | Mixed |
| use Ours instead of LineDL | RAW denoising ELD x100 · ELD RAW denoising x100 | PSNR: 43.34 -> 44.38 | Improved |
Closest papers by meaning
- Memory-Efficient Hierarchical Neural Architecture Search for Image Restoration (IJCV 2021 · similarity 0.84)
- Memory-Efficient Hierarchical Neural Architecture Search for Image Denoising (CVPR 2020 · similarity 0.84)
- Fast Image Processing With Fully-Convolutional Networks (ICCV 2017 · similarity 0.83)
- Unprocessing Images for Learned Raw Denoising (CVPR 2019 · similarity 0.83)
- Adaptive Dynamic Filtering Network for Image Denoising (AAAI 2023 · similarity 0.83)
- Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional Neural Networks (AAAI 2024 · 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.
