Research map / Video / Video segmentation and tracking
High-quality Single-model Deep Video Compression with Frame-Conv3D and Multi-frame Differential Modulation
Wenyu Sun, Chen Tang, Weigui Li, Zhuqing Yuan, Huazhong Yang, Yongpan Liu
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 |
|---|---|---|---|
| replace Channel-Conv3D with Frame-Conv3D (ablation) | video codec with traditional quantizer · UVG, half of frames dropped | average PSNR (dB): 24.94 -> 26.37 | Improved |
| add differential quantizer in place of traditional quantizer (ablation) | Frame-Conv3D codec · UVG, half of frames dropped | average PSNR (dB): 26.37 -> 29.40 | Improved |
| add differential quantizer in place of traditional quantizer (ablation) | Frame-Conv3D codec · validation subset, 0.25 BPP training | PSNR (dB): >10 dB | Improved |
| add differential quantizer in place of traditional quantizer (ablation) | Frame-Conv3D codec · entropy of bit streams, bit width 16 | entropy coding gain (%): nearly 30% | Improved |
| use 2nd order instead of 1st order differential modulation (ablation) | differential quantizer · UVG | PSNR and MS-SSIM | No clear effect |
| add prediction network (ablation) | video codec · UVG | PSNR and MS-SSIM | Improved |
| add prediction loss d(F, F-bar) with weight 0.05 (ablation) | video codec · UVG | PSNR | Improved |
| code 4 frames instead of 8 (ablation) | video codec · UVG validation subset, 0.25 BPP | convergence and final PSNR | No clear effect |
| down-scale to 1/4 (M=2) instead of 1/8 (M=3) (ablation) | video codec · UVG validation subset, 0.25 BPP | PSNR | Improved |
| add dropout scheme during training with 50%, 75%, 87.5% frame dropping (ablation) | differential quantizer codec · UVG, variable BPP | PSNR at different bit rates | Improved |
| compare proposed codec with H.264 and H.265 at low BPP | proposed codec · UVG, HEVC Class D, Class E | PSNR and MS-SSIM | Improved |
| evaluate on fast-motion sequences | proposed codec · HEVC Class B and C | PSNR and MS-SSIM | Hurt |
Closest papers by meaning
- Learning for Video Compression With Hierarchical Quality and Recurrent Enhancement (CVPR 2020 · similarity 0.85)
- DVC: An End-To-End Deep Video Compression Framework (CVPR 2019 · similarity 0.85)
- C3: High-Performance and Low-Complexity Neural Compression from a Single Image or Video (CVPR 2024 · similarity 0.83)
- 3D-LMVIC: Learning-based Multi-View Image Compression with 3D Gaussian Geometric Priors (ICML 2025 · similarity 0.83)
- High Dynamic Range Video Compression: A Large-Scale Benchmark Dataset and A Learned Bit-depth Scalable Compression Algorithm (CVPR 2025 · similarity 0.83)
- Neural Video Compression with Feature Modulation (CVPR 2024 · 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.
