Research map / Image generation / Image restoration
Edit-aware RAW reconstruction
Abhijith Punnappurath, Luxi Zhao, Ke Zhao, Hue Nguyen, Radek Grzeszczuk, Michael S. Brown
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 edit-aware loss | CAM · Edit 1 sRGB | PSNR: 27.27 -> 29.24 | Improved |
| add edit-aware loss | RAW Diffusion · Edit 5 sRGB | PSNR: 23.29 -> 25.03 | Improved |
| add edit-aware loss | UNet · Edit 5 sRGB | PSNR: 26.44 -> 28.02 | Improved |
| add edit-aware loss (without RAW loss) | UNet · RAW reconstruction | PSNR: 38.82 -> 35.62 | Hurt |
| add edit-aware loss (without RAW loss) | RAW Diffusion · RAW reconstruction | PSNR: 34.18 -> 33.37 | Hurt |
| add edit-aware loss | CAM · RAW reconstruction | PSNR: 37.17 -> 37.57 | Improved |
| add edit-aware loss | CAM · Image-specific edits (50 hard images) | PSNR: 22.11 -> 24.36 | Improved |
| add edit-aware fine-tuning on the target edit | UNet · EV +2 fine-tuning | PSNR: 31.31 -> 31.51 | Improved |
| use all four loss modules instead of exposure only (ablation) | UNet · Edit 5 sRGB (50 hard images) | PSNR: 23.22 -> 25.15 | Improved |
| use all four loss modules instead of color only (ablation) | UNet · Edit 5 sRGB (50 hard images) | PSNR: 20.54 -> 25.15 | Improved |
| add randomized ISP parameter sampling (ablation) | UNet · Edit 5 sRGB (50 hard images) | PSNR: 24.20 -> 25.15 | Improved |
| use edit-aware loss instead of cyclic consistency loss (ablation) | CIE XYZ Net · Edit 5 sRGB (50 hard images) | PSNR: 20.54 -> 27.13 | Improved |
Closest papers by meaning
- Learning sRGB-to-Raw-RGB De-Rendering With Content-Aware Metadata (CVPR 2022 · similarity 0.88)
- RAW Image Reconstruction Using a Self-Contained sRGB-JPEG Image With Only 64 KB Overhead (CVPR 2016 · similarity 0.88)
- Raw-JPEG Adapter: Efficient Raw Image Compression with JPEG (ECCV 2026 · similarity 0.88)
- Leveraging Frame Affinity for sRGB-to-RAW Video De-rendering (CVPR 2024 · similarity 0.87)
- RAW Image Reconstruction Using a Self-contained sRGB–JPEG Image with Small Memory Overhead (IJCV 2017 · similarity 0.87)
- CODE: Confident Ordinary Differential Editing (AAAI 2025 · 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.
