Research map / Video / Video segmentation and tracking
Video Compression through Image Interpolation
Chao-Yuan Wu, Nayan Singhal, Philipp Krahenbuhl
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 entropy coding (3D PixelCNN with adaptive arithmetic coding) (ablation) | All interpolation models · VTL | BPP: saves up to 52% BPP at low bitrate, at least 10% at high bitrate | Improved |
| Add motion compensation (block motion vectors) to interpolation (ablation) | Interpolation network · VTL | MS-SSIM vs BPP | Improved |
| Use block motion vectors instead of optical flow (ablation) | M6,6 model · VTL | MS-SSIM vs BPP | Improved |
| Assume lossless free flow compression (flow upper bound) vs motion vectors (ablation) | M6,6 model · VTL | MS-SSIM vs BPP | Improved |
| Add residual encoding to motion compensated interpolation (ablation) | Interpolation network · VTL | MS-SSIM vs BPP | Improved |
| Use end-to-end interpolation instead of deep image compression of traditional residuals (ablation) | Residual baseline · VTL | MS-SSIM vs BPP | Improved |
| Extend binary hierarchy to n=27 frames at all levels (ablation) | Hierarchical interpolation · VTL | performance | Hurt |
| Use more than three hierarchy levels (ablation) | Hierarchical interpolation · VTL | performance | Hurt |
| Halve U-net channel counts (ablation) | Context model · VTL | performance | No clear effect |
| Fuse U-net features at more resolutions (ablation) | Interpolation network · VTL | performance | Improved |
Closest papers by meaning
- Neural Inter-Frame Compression for Video Coding (ICCV 2019 · similarity 0.89)
- An End-to-End Learning Framework for Video Compression (TPAMI 2020 · similarity 0.88)
- DVC: An End-To-End Deep Video Compression Framework (CVPR 2019 · similarity 0.88)
- Learned Video Compression (ICCV 2019 · similarity 0.87)
- What Matters in Practical Learned Image Compression (CVPR 2026 · similarity 0.87)
- Conditional Entropy Coding for Efficient Video Compression (ECCV 2020 · 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.
