Measuring Video Quality Using AVQT
summary
MeAVQT is a macOS command-line tool that estimates perceptual quality by comparing a reference and a compressed video using AVFoundation for formats and Metal for GPU processing. It supports SDR and HDR inputs, requires spatial and temporal alignment (a --force flag bypasses this), accepts display resolution and viewing distance parameters, and reports frame and segment scores. MeAVQT achieves ~90% correlation with human scores on public datasets and runs >170 fps on an M1 Mac; a Linux port is in progress.
AI-written summary · about
chapters (10)
- 0:00AVQT introduction and challenges
- 1:31Perceptual quality importance
- 2:38AVQT features
- 3:21Perceptual accuracy by content
- 5:48AVQT on public datasets
- 7:34AVQT speed and scalability
- 9:02Viewing setup awareness
- 10:27AVQT demo and CLI
- 12:55AVQT takeaways and future
- 14:00Q&A
related talks
05mVideo Engineer Toolbox
53mQt Multimedia
22mKeynote: WTF Is AQI Ad Quality Index
21mQuickCut: An Interactive Tool for Editing Narrated Video
19mVidCrit: Video-based Asynchronous Video Review
15mMeasuring developer productivity through their own product measurements
01h 00mMeasuring Qt Performance
38mImprove your app and game quality with Android vitals