Research map / 3D vision / Depth and stereo
Single-Image HDR Reconstruction by Learning to Reverse the Camera Pipeline
Check 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 |
|---|---|---|---|
| Fine-tune the model on HDR-Real | Ours* · HDR-Real | HDR-VDP-2: 51.59±7.42 -> 53.16±7.19 | Helped |
| Fine-tune the model on HDR-Real | Ours* · RAISE | HDR-VDP-2: 58.80±3.91 -> 59.21±3.68 | Helped |
| Fine-tune the model on HDR-Real | Ours* · HDR-Eye | HDR-VDP-2: 52.66±5.64 -> 53.16±5.92 | Helped |
| Add Dequantization-Net | HDR image dequantization · HDR-Synth dequantization | PSNR: 33.86±6.96 -> 35.87±6.11 | Helped |
| Add edge features (ablation) | Linearization-Net · HDR-Synth linearization | L2 error of inverse CRF: 2.00±3.15 -> 1.66±2.93 | Helped |
| Add histogram features (ablation) | Linearization-Net · HDR-Synth linearization | L2 error of inverse CRF: 2.00±3.15 -> 1.61±3.03 | Helped |
| Add edge and histogram features (ablation) | Linearization-Net · HDR-Synth linearization | PSNR of linear image: 33.43±7.03 -> 34.53±6.83 | Helped |
| Enforce a monotonically increasing inverse CRF (ablation) | Linearization-Net with image, edge and histogram features · HDR-Synth linearization | L2 error of inverse CRF: 1.58±2.73 -> 1.56±2.52 | Helped |
| Enforce a monotonically increasing inverse CRF (ablation) | Linearization-Net with image, edge and histogram features · HDR-Synth linearization | PSNR of linear image: 34.53±6.83 -> 34.64±6.73 | Helped |
| Add positive-residual constraint (ablation) | Hallucination-Net · HDR-Synth | HDR-VDP-2: 63.60±15.32 -> 64.79±15.89 | Helped |
| Replace transposed convolution with resize convolution (ablation) | Hallucination-Net with positive residuals · HDR-Synth | HDR-VDP-2: 64.79±15.89 -> 64.52±16.05 | Hurt |
| Add perceptual loss (ablation) | Hallucination-Net with resize convolution and positive residuals · HDR-Synth | HDR-VDP-2: 64.52±16.05 -> 66.31±15.82 | Helped |
Closest papers by meaning
- Single Shot HDR Recovery via a Video Diffusion Prior (NeurIPS 2026 · similarity 0.90)
- Towards Real-World HDR Video Reconstruction: A Large-Scale Benchmark Dataset and A Two-Stage Alignment Network (CVPR 2024 · similarity 0.89)
- Self-Supervised High Dynamic Range Imaging with Multi-Exposure Images in Dynamic Scenes (ICLR 2024 · similarity 0.88)
- High Dynamic Range Novel View Synthesis with Single Exposure (ICML 2025 · similarity 0.88)
- Generating Content for HDR Deghosting from Frequency View (CVPR 2024 · similarity 0.88)
- Zero-Shot Structure-Preserving Diffusion Model for High Dynamic Range Tone Mapping (CVPR 2024 · similarity 0.87)
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.