Research map / Video / Video segmentation and tracking
Extending Neural P-Frame Codecs for B-Frame Coding
Reza Pourreza, Taco Cohen
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 frame interpolation of two references so a P-frame codec codes B-frames (B-EPIC vs SSF) | SSF P-frame codec · UVG | PSNR BD-rate gain vs H.264 (%): -25.71 -> -47.89 | Improved |
| add frame interpolation of two references so a P-frame codec codes B-frames (B-EPIC vs SSF) | SSF P-frame codec · UVG | MS-SSIM BD-rate gain vs H.264 (%): -41.05 -> -50.79 | Improved |
| add frame interpolation of two references so a P-frame codec codes B-frames (B-EPIC vs SSF) | SSF P-frame codec · MCL-JCV | PSNR BD-rate gain vs H.264 (%): -15.90 -> -31.91 | Improved |
| add frame interpolation of two references so a P-frame codec codes B-frames (B-EPIC vs SSF) | SSF P-frame codec · HEVC-C | PSNR BD-rate gain vs H.264 (%): 47.69 -> 11.22 | Improved |
| add frame interpolation of two references so a P-frame codec codes B-frames (B-EPIC vs SSF) | SSF P-frame codec · HEVC-D | MS-SSIM BD-rate gain vs H.264 (%): -20.79 -> -43.98 | Improved |
| B-EPIC (MSE) vs H.265 in PSNR on small-resolution HEVC-D (padding to multiple of 64) | B-EPIC (MSE) · HEVC-D | PSNR BD-rate gain vs H.264 (%): -16.25 -> 29.05 | Hurt |
| use FFMPEG default parameters for H.264/H.265 anchors instead of GoP=12 settings | B-EPIC (MSE) · UVG | PSNR BD-rate gain vs H.264 (%): -47.89 -> -26.93 | Hurt |
| train on longer or sparser sequences (7 consecutive frames, or 4 frames two apart) instead of 4 consecutive frames (ablation) | B-EPIC · UVG | rate-distortion | No clear effect |
| increase training GoP length (ablation) | B-EPIC · Vimeo-90k training | rate-distortion | No clear effect |
Closest papers by meaning
- Neural B-frame Video Compression with Bi-directional Reference Harmonization (NeurIPS 2025 · similarity 0.86)
- Neural Video Compression with Context Modulation (CVPR 2025 · similarity 0.85)
- PNVC: Towards Practical INR-based Video Compression (AAAI 2025 · similarity 0.84)
- Neural Video Compression with Feature Modulation (CVPR 2024 · similarity 0.84)
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.83)
- Perceptual Video Compression with Neural Wrapping (CVPR 2025 · similarity 0.83)
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.
