Research map / Video / Video segmentation and tracking
Hybrid Neural Fusion for Full-Frame Video Stabilization
Yu-Lun Liu, Wei-Sheng Lai, Ming-Hsuan Yang, Yung-Yu Chuang, Jia-Bin Huang
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 |
|---|---|---|---|
| fuse in hybrid space instead of feature space (ablation) | CNN-based fusion · Fusion ablation (940 synthetic test videos) | PSNR: 27.187 -> 27.868 | Improved |
| fuse in hybrid space instead of image space (ablation) | CNN-based fusion · Fusion ablation (940 synthetic test videos) | PSNR: 26.692 -> 27.868 | Improved |
| fuse in hybrid space instead of feature space (ablation) | CNN-based fusion · Fusion ablation (940 synthetic test videos) | LPIPS: 0.092 -> 0.073 | Improved |
| use CNN-based fusion function instead of mean in hybrid space (ablation) | hybrid-space fusion · Fusion ablation (940 synthetic test videos) | PSNR: 27.013 -> 27.868 | Improved |
| use CNN-based fusion function instead of argmax in hybrid space (ablation) | hybrid-space fusion · Fusion ablation (940 synthetic test videos) | PSNR: 27.519 -> 27.868 | Improved |
| use CNN-based fusion instead of argmax in image space (ablation) | image-space fusion · Fusion ablation (940 synthetic test videos) | SSIM: 0.928 -> 0.895 | Hurt |
| add flow-error weighting instead of mean in image space (ablation) | image-space fusion · Fusion ablation (940 synthetic test videos) | PSNR: 24.618 -> 25.363 | Improved |
| combine proposed method with Bundle smoothing | Bundle · Selfie dataset | accumulated optical flow: 0.84 -> 0.70 | Improved |
| combine proposed method with Yu and Ramamoorthi smoothing | Yu and Ramamoorthi · Selfie dataset | cropping ratio: 0.78 -> 1.00 | Improved |
| combine proposed method with Bundle smoothing | Bundle · Selfie dataset | stability score: 0.85 -> 0.85 | No clear effect |
| use hybrid fusion for view synthesis instead of FVS | FVS · Tanks and Temples Truck | PSNR: 22.21 -> 23.25 | Improved |
| full-frame fusion instead of cropping-based stabilizers | Adobe Premiere Pro · NUS dataset | cropping ratio: 0.74 -> 1.00 | Improved |
Closest papers by meaning
- Fast Full-frame Video Stabilization with Iterative Optimization (ICCV 2023 · similarity 0.90)
- Learning Video Stabilization Using Optical Flow (CVPR 2020 · similarity 0.89)
- 3D Video Stabilization With Depth Estimation by CNN-Based Optimization (CVPR 2021 · similarity 0.88)
- Harnessing Meta-Learning for Improving Full-Frame Video Stabilization (CVPR 2024 · similarity 0.88)
- 3D Multi-frame Fusion for Video Stabilization (CVPR 2024 · similarity 0.85)
- A Unified Solution to Video Fusion: From Multi-Frame Learning to Benchmarking (NeurIPS 2025 · similarity 0.85)
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.
