Research map / Image generation / Super-resolution
Dual-Camera Super-Resolution With Aligned Attention Modules
Tengfei Wang, Jiaxin Xie, Wenxiu Sun, Qiong Yan, Qifeng Chen
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 |
|---|---|---|---|
| use learnable confidence map (adaptive fusion) instead of the original confidence map (ablation) | Ours-l1 · CUFED5 | PSNR: 27.12 -> 27.30 | Improved |
| use adaptive fusion instead of element-wise summation (ablation) | Ours-l1 · CUFED5 | PSNR: 26.85 -> 27.30 | Improved |
| use adaptive fusion instead of element-wise summation (ablation) | Ours-l1 · CUFED5 | SSIM: 0.794 -> 0.807 | Improved |
| add aligned attention (local patch-wise affine alignment) to patch matching (ablation) | Ours · CUFED5 | warped reference quality | Improved |
| add fidelity loss (ablation) | Ours · CUFED5 | visual detail | Improved |
| add high-frequency residual fusion in image space (ablation) | Ours · CUFED5 | visual detail | Improved |
| add self-supervised real-image adaptation (ablation) | Ours · CameraFusion (real full-resolution) | visual sharpness | Improved |
| retrain with l1 loss only instead of perceptual/adversarial losses | Ours · CUFED5 | PSNR: 25.39 -> 27.30 | Improved |
| use less similar reference images (L1 to L4) (ablation) | Ours-l1 · CUFED5 | PSNR: 27.30 -> 26.70 | Hurt |
| use less similar reference images (L1 to L4) | TTSR-l1 · CUFED5 | PSNR: 26.99 -> 26.58 | Hurt |
| evaluate outside the overlapped FoV where no reference patches exist | Ours-l1 · CameraFusion | PSNR: 37.28 -> 36.94 | No clear effect |
Closest papers by meaning
- Self-Supervised Learning for Real-World Super-Resolution From Dual and Multiple Zoomed Observations (TPAMI 2024 · similarity 0.93)
- Self-Supervised Learning for Real-World Super-Resolution from Dual Zoomed Observations (ECCV 2022 · similarity 0.92)
- RRSR:Reciprocal Reference-Based Image Super-Resolution with Progressive Feature Alignment and Selection (ECCV 2022 · similarity 0.90)
- LMR: A Large-Scale Multi-Reference Dataset for Reference-Based Super-Resolution (ICCV 2023 · similarity 0.89)
- Reference-based Image and Video Super-Resolution via $C^{2}$-Matching (TPAMI 2022 · similarity 0.89)
- KeDuSR: Real-World Dual-Lens Super-Resolution via Kernel-Free Matching (AAAI 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.
