Research map / Video / Video segmentation and tracking
Efficient Video Compression via Content-Adaptive Super-Resolution
Mehrdad Khani, Vibhaalakshmi Sivaraman, Mohammad Alizadeh
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 periodic model adaptation instead of one-shot customization (ablation) | SRVC · Meridian video | PSNR (dB): 35.32 -> 37.25 | Improved |
| update the model every 5 s instead of every 10 s (ablation) | SRVC · Meridian video | PSNR (dB): 36.52 -> 37.25 | Improved |
| update the model every 20 s instead of 5 s (ablation) | SRVC · Meridian video | PSNR (dB): 37.25 -> 36.45 | Hurt |
| increase output feature channels from 16 to 32 (ablation) | SRVC · Video snippet | PSNR (dB): 38.69 -> 39.87 | Improved |
| increase output feature channels from 32 to 128 (ablation) | SRVC · Video snippet | PSNR (dB): 39.87 -> 39.90 | No clear effect |
| increase output feature channels from 32 to 64 (ablation) | SRVC · Video snippet | inference time (ms): 11 -> 17 | Hurt |
| raise content stream quality from CRF 35 to CRF 20 (ablation) | SRVC · Meridian video | PSNR (dB): 31 dB to 36 dB | Improved |
| raise content bitrate further beyond CRF 20 (ablation) | SRVC · Meridian video | PSNR (dB) | No clear effect |
| adapt 10% of model parameters instead of 0.5% (ablation) | SRVC · Meridian video | PSNR (dB): 36.31 dB to 37.32 dB | Improved |
| use one-shot customized model instead of bicubic upsampling (ablation) | SRVC lightweight model · Xiph | PSNR | Hurt |
| use generic EDSR super-resolution instead of SRVC | 480p H.265 + SR · Xiph | PSNR vs bpp | No clear effect |
| add content-adaptive SR to a 480p H.265 stream versus H.265 at 1080p | SRVC · Xiph | bits-per-pixel at equal PSNR: 0.03 -> 0.005 | Improved |
Closest papers by meaning
- A Codec Information Assisted Framework for Efficient Compressed Video Super-Resolution (ECCV 2022 · similarity 0.88)
- DVC: An End-To-End Deep Video Compression Framework (CVPR 2019 · similarity 0.87)
- Learning for Video Compression With Hierarchical Quality and Recurrent Enhancement (CVPR 2020 · similarity 0.87)
- Towards Practical Real-Time Neural Video Compression (CVPR 2025 · similarity 0.87)
- SAVSR: Arbitrary-Scale Video Super-Resolution via a Learned Scale-Adaptive Network (AAAI 2024 · similarity 0.86)
- PNVC: Towards Practical INR-based Video Compression (AAAI 2025 · similarity 0.86)
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.
