Research map / Video / Video segmentation and tracking
M-LVC: Multiple Frames Prediction for Learned Video Compression
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 MAMVP-Net to predict motion vectors from multiple prior MV fields (ablation) | Our Baseline · HEVC Class D | PSNR: about 0.5~0.7 dB | Helped |
| Add MV refinement after MAMVP-Net (ablation) | Add MAMVP-Net · HEVC Class D | PSNR: about 0.15dB at high bit rates and about 0.4dB at low bit rates | Helped |
| Add the previous reconstructed frame as an MV Refine-Net input (ablation) | Add MVRefine-Net-0 · HEVC Class D | PSNR: about 0.1dB | Helped |
| Use multiple reference frames in motion compensation (ablation) | Add MVRefine-Net · HEVC Class D | PSNR: about 0.1~0.25dB | Helped |
| Add residual refinement (ablation) | Add MMC-Net · HEVC Class D | PSNR: about 0.3dB gain at low bit rates and about 0.2dB gain at high bit rates | Helped |
| Train the proposed network from scratch instead of using step-by-step training (ablation) | Proposed · Kimono | bitrate allocation: 0.0002bpp for MVD and 0.2431bpp for residual | Hurt |
| Use multi-scale aligned MV prediction rather than single-scale w/o alignment (ablation) | MAMVP-Net · HEVC Class D | PSNR (dB): 31.250 -> 31.290 | Helped |
| Use multi-scale aligned MV prediction rather than single-scale w/o alignment (ablation) | MAMVP-Net · HEVC Class D | bpp: 0.297 -> 0.285 | Helped |
| Replace single-reference motion compensation with multiple-reference MMC-Net (ablation) | Add MVRefine-Net · BasketballPass, frame 9 | PSNR of predicted frame: 31.42dB -> 31.97dB | Helped |
| Increase the number of reference frames from two to three or four (ablation) | M-LVC · HEVC Class D | compression efficiency | Helped |
Closest papers by meaning
- ECVC: Exploiting Non-Local Correlations in Multiple Frames for Contextual Video Compression (CVPR 2025 · similarity 0.88)
- Neural Video Compression with Context Modulation (CVPR 2025 · similarity 0.86)
- FLAVC: Learned Video Compression with Feature Level Attention (CVPR 2025 · similarity 0.86)
- Neural B-frame Video Compression with Bi-directional Reference Harmonization (NeurIPS 2025 · similarity 0.85)
- Deep Hierarchical Video Compression (AAAI 2024 · similarity 0.85)
- Neural Video Compression with Feature Modulation (CVPR 2024 · 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.