Research map / 3D vision / Depth and stereo
Neural Image Representations for Multi-Image Fusion and Layer Separation
Seonghyeon Nam, Marcus A. Brubaker, Michael S. Brown
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 |
|---|---|---|---|
| learn homography jointly with the layer separation via an MLP instead of conventional homography estimation (ablation) | homography-based NIR · Synthetic moire removal | PSNR / SSIM: 23.94 -> 38.68 PSNR; 0.7690 -> 0.9751 SSIM | Improved |
| use a simple homography motion model instead of the more general occlusion-aware flow model (ablation) | NIR for demoireing · Synthetic moire removal | PSNR / SSIM: 36.72 -> 38.68 PSNR; 0.9512 -> 0.9751 SSIM | Improved |
| add the interference loss L_Interf (ablation) | reflection-removal NIR · Real reflection images | qualitative | Improved |
| add the flow smoothness loss L_TVFlow (ablation) | reflection-removal NIR · Real reflection images | qualitative | Improved |
| add the exclusion loss L_Excl (ablation) | reflection-removal NIR · Real reflection images | qualitative | Improved |
| add the extra dimension w for occluded appearance (ablation) | occlusion-aware flow NIR · RainSynLight25 | qualitative | Improved |
| add L1 regularization on w to center it at zero (ablation) | occlusion-aware flow NIR · RainSynLight25 | PSNR | No clear effect |
| use more input images (ablation) | NIR rain removal · Rain removal | qualitative | Improved |
| unsupervised NIR versus supervised AFN | NIR · Synthetic moire removal | PSNR: 43.63 -> 38.68 | Hurt |
| unsupervised NIR versus supervised FCDN | NIR · RainSynLight25 | PSNR: 35.80 -> 28.61 | Hurt |
Closest papers by meaning
- Neural Spline Fields for Burst Image Fusion and Layer Separation (CVPR 2024 · similarity 0.86)
- Neural Experts: Mixture of Experts for Implicit Neural Representations (NeurIPS 2024 · similarity 0.84)
- Explaining the Implicit Neural Canvas: Connecting Pixels to Neurons by Tracing their Contributions (CVPR 2024 · similarity 0.84)
- Neural Assets: 3D-Aware Multi-Object Scene Synthesis with Image Diffusion Models (NeurIPS 2024 · similarity 0.84)
- Flow-NeRF: Joint Learning of Geometry, Poses, and Dense Flow within Unified Neural Representations (CVPR 2025 · similarity 0.84)
- PM-INR: Prior-Rich Multi-Modal Implicit Large-Scale Scene Neural Representation (AAAI 2024 · similarity 0.84)
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.
