Research map / Image generation / Super-resolution
Towards Practical and Efficient High-Resolution HDR Deghosting with CNN
K. Ram Prabhakar, Susmit Agrawal, Durgesh Kumar Singh, Balraj Ashwath, R. Venkatesh Babu
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 Nmc motion segmentation together with Ng weight-map fusion for the guide image (ablation) | guide image Hg · UCSD test | PSNR-T: 37.24 -> 41.56 | Improved |
| replace simple difference-based motion detection with learned Nmc before Ng (ablation) | guide image Hg · UCSD test | PSNR-T: 39.06 -> 41.56 | Improved |
| add Ng learned weight maps instead of triangle fusion after Nmc (ablation) | guide image Hg · UCSD test | PSNR-T: 37.98 -> 41.56 | Improved |
| add optical flow alignment of low-resolution inputs to Nf (ablation) | Nf · UCSD test (d=1) | PSNR-T: 41.16 -> 42.82 | Improved |
| use optical-flow aligned inputs instead of motion-compensated sequence for Nf (ablation) | Nf · UCSD test (d=1) | PSNR-T: 41.54 -> 42.82 | Improved |
| add XSDC blocks instead of ten vanilla convolutional layers (ablation) | Nf · UCSD test | PSNR-T: 41.08 -> 42.82 | Improved |
| extend SDC to XSDC (three consecutive convs, 1x1 aggregation) (ablation) | Nf · UCSD test | PSNR-T: 41.94 -> 42.34 | Improved |
| add dense connections among XSDC blocks (ablation) | Nf · UCSD test | PSNR-T: 42.34 -> 42.82 | Improved |
| use XSDC instead of ResNet18 backbone (ablation) | Nf · UCSD test | PSNR-T: 42.45 -> 42.82 | Improved |
| add BGU low-resolution processing (d=8) instead of full-resolution processing (d=1) (ablation) | Ours · Kalantari UCSD test | PSNR-T: 43.08 -> 42.82 | Hurt |
| use low-resolution processing with BGU for high-resolution images | Ours · 16 MP images | running time: 11 seconds for 16 MP vs 42 seconds for 1 MP (Kalantari17) | Improved |
Closest papers by meaning
- A Unified HDR Imaging Method With Pixel and Patch Level (CVPR 2023 · similarity 0.92)
- Ghost-Free High Dynamic Range Imaging with Context-Aware Transformer (ECCV 2022 · similarity 0.91)
- Generating Content for HDR Deghosting from Frequency View (CVPR 2024 · similarity 0.89)
- HDR-GS: Efficient High Dynamic Range Novel View Synthesis at 1000x Speed via Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- Self-Supervised High Dynamic Range Imaging with Multi-Exposure Images in Dynamic Scenes (ICLR 2024 · similarity 0.87)
- Single Shot HDR Recovery via a Video Diffusion Prior (NeurIPS 2026 · 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.
