Research map / Image generation / Image restoration
Model-Based Image Signal Processors via Learnable Dictionaries
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the learnable dictionary-based ISP instead of CycleISP for RAW reconstruction | RAW reconstruction · SIDD validation | PSNR_r (dB): 37.62 -> 45.21; +7.6 dB | Helped |
| use the learnable dictionary-based ISP instead of U-Net for RAW reconstruction | RAW reconstruction · SIDD validation | PSNR_r (dB): 39.84 -> 45.21 | Helped |
| use the learnable dictionary-based ISP instead of Invertible-ISP with JPEG Fourier | RAW reconstruction · Canon EOS 5D | PSNR_r: 46.78 -> 50.08 | Helped |
| use the learnable dictionary-based ISP instead of Invertible-ISP with JPEG Fourier | RAW reconstruction · Nikon D700 | PSNR_r: 44.42 -> 43.62 | Hurt |
| train the denoiser on synthetic RAW data generated by the proposed ISP | U-Net denoiser · SIDD | PSNR_d (dB): +0.7 dB | Helped |
| add dictionary-based data augmentation | DHDN denoiser · SIDD | PSNR: +0.46 dB | Helped |
| train using few clean pairs and dictionary augmentation (ablation) | DHDN denoiser (Ours-f) · SIDD | PSNR: 52.02 -> 52.05 | Mixed |
| remove clean ground-truth denoising labels (ablation) | DHDN denoiser (Ours-u) · SIDD | PSNR: 52.48 -> 49.90 | Hurt |
| apply learned color correction (ablation) | Proposed ISP · SIDD | Chrominance UV PSNR: 36.21dB -> 43.34dB | Helped |
| apply learned white-balance gains (ablation) | Proposed ISP · SIDD | Chrominance UV PSNR: 43.34dB -> 47.59dB | Helped |
| apply lens-shading-effect modelling (ablation) | Proposed ISP · SIDD | Luminance PSNR: 39.91dB -> 51.24dB | Helped |
| use the full proposed denoising pipeline instead of CycleISP | RAW denoising · SIDD | PSNR: 52.38 -> 52.48 | Helped |
Closest papers by meaning
- ISPDiffuser: Learning RAW-to-sRGB Mappings with Texture-Aware Diffusion Models and Histogram-Guided Color Consistency (AAAI 2025 · similarity 0.87)
- ReRAW: RGB-to-RAW Image Reconstruction via Stratified Sampling for Efficient Object Detection on the Edge (CVPR 2025 · similarity 0.86)
- DiffRAW: Leveraging Diffusion Model to Generate DSLR-Comparable Perceptual Quality sRGB from Smartphone RAW Images (AAAI 2024 · similarity 0.86)
- RAWild: Toward Sensor-Agnostic RAW Object Detection via Physics-Guided Curve and Grid Modeling (NeurIPS 2026 · similarity 0.86)
- Enhancing RAW-to-sRGB with Decoupled Style Structure in Fourier Domain (AAAI 2024 · similarity 0.86)
- RGB-Only Supervised Camera Parameter Optimization in Dynamic Scenes (NeurIPS 2025 · similarity 0.86)
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.