Research map / Image generation / Image restoration
Beyond Learned Metadata-Based Raw Image Reconstruction
Yufei Wang, Yi Yu, Wenhan Yang, Lanqing Guo, Lap‐Pui Chau, Alex Chichung Kot, Bihan Wen
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 the proposed sRGB-guided context model (ablation) | metadata codec · NUS Sony | bpp: 0.5621 -> 0.2619 | Improved |
| add the proposed sRGB-guided context model (ablation) | metadata codec · NUS Olympus | bpp: 1.0411 -> 0.4115 | Improved |
| add the proposed sRGB-guided context model (ablation) | metadata codec · NUS Samsung | bpp: 0.8368 -> 0.4555 | Improved |
| use predicted encoding order instead of reversed order (ablation) | metadata codec · NUS Sony | bpp: 0.3740 -> 0.2619 | Improved |
| use the proposed context model instead of the one from prior work (ablation) | metadata codec · NUS Sony | bpp: 0.7823 -> 0.2619 | Improved |
| use the proposed context model instead of the one from prior work (ablation) | metadata codec · NUS Sony | PSNR (dB): 60.35 -> 58.21 | Hurt |
| use asymmetric hybrid backbone instead of resolution-maintained backbone (ablation) | reconstruction backbone · NUS Sony | bpp: 1.36 -> 0.26 | Improved |
| use asymmetric hybrid backbone instead of resolution-maintained backbone (ablation) | reconstruction backbone · NUS Sony | PSNR (dB): 55.04 -> 58.21 | Improved |
| use asymmetric hybrid backbone instead of x16 downsampling (ablation) | reconstruction backbone · NUS Sony | PSNR (dB): 48.37 -> 58.21 | Improved |
| add adaptive quantization bin width (ablation) | metadata codec · NUS | rate-distortion | Improved |
| use metadata with the same backbone instead of no metadata | raw reconstruction · AdobeFiveK software ISP | PSNR (dB): 53.03 -> 58.44 | Improved |
| use this work instead of conference version R2LCM at low rate | raw reconstruction · AdobeFiveK software ISP | PSNR (dB): 58.14 -> 58.44 | Improved |
Closest papers by meaning
- Raw Image Reconstruction With Learned Compact Metadata (CVPR 2023 · similarity 0.96)
- Metadata-Based RAW Reconstruction via Implicit Neural Functions (CVPR 2023 · similarity 0.91)
- Learning Raw Image Reconstruction-Aware Deep Image Compressors (TPAMI 2019 · similarity 0.90)
- Compress Clean Signal from Noisy Raw Image: A Self-Supervised Approach (ICML 2024 · similarity 0.87)
- Leveraging Frame Affinity for sRGB-to-RAW Video De-rendering (CVPR 2024 · similarity 0.87)
- Ultra Lowrate Image Compression with Semantic Residual Coding and Compression-aware Diffusion (ICML 2025 · similarity 0.85)
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.
