Research map / Image generation / Image restoration
Learning RAW-to-sRGB Mappings With Inaccurately Aligned Supervision
Zhilu Zhang, Haolin Wang, Ming Liu, Ruohao Wang, Jiawei Zhang, Wangmeng Zuo
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 GCM-guided optical-flow alignment (target aligned to GCM output) (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.49 -> 23.76 | Improved |
| align target with network output instead of GCM output (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.76 -> 23.33 | Hurt |
| align target with demosaicked raw instead of GCM output (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.76 -> 23.52 | Hurt |
| add SPN to GCM-free baseline (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.49 -> 23.61 (0.12 dB gain) | Improved |
| add target sRGB guidance to SPN (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.61 -> 23.69 | Improved |
| add coordinate map to SPN + y (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.69 -> 23.76 | Improved |
| increase RCABs from 2 to 4 (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.54 -> 23.76 | Improved |
| increase RCABs from 4 to 8 (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.76 -> 23.74 | No clear effect |
| increase RCABs from 4 to 20 (ablation) | LiteISPNet · ZRR (Align GT with raw) | PSNR: 23.76 -> 23.79 (23.3 ms -> 91.5 ms) | No clear effect |
| replace MW-ISPNet with LiteISPNet | MW-ISPNet · ZRR (Align GT with raw) | PSNR: 23.07 -> 23.76 | Improved |
| replace MW-ISPNet (GAN) with LiteISPGAN | MW-ISPNet (GAN) · ZRR (Align GT with raw) | LPIPS: 0.134 -> 0.120 | Improved |
| use joint alignment training instead of Zhang et al. CoBi loss | SRResNet · SR-RAW 35/150 (Align GT with result) | PSNR: 22.26 -> 22.50 | Improved |
Closest papers by meaning
- Learning sRGB-to-Raw-RGB De-Rendering With Content-Aware Metadata (CVPR 2022 · similarity 0.87)
- ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency (AAAI 2025 · similarity 0.87)
- Enhancing RAW-to-sRGB with Decoupled Style Structure in Fourier Domain (AAAI 2024 · similarity 0.86)
- DiffRAW: Leveraging Diffusion Model to Generate DSLR-Comparable Perceptual Quality sRGB from Smartphone RAW Images (AAAI 2024 · similarity 0.85)
- ReRAW: RGB-to-RAW Image Reconstruction via Stratified Sampling for Efficient Object Detection on the Edge (CVPR 2025 · similarity 0.84)
- ThermalGen: Style-Disentangled Flow-Based Generative Models for RGB-to-Thermal Image Translation (NeurIPS 2025 · similarity 0.83)
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.
