Schémas hybrides pour la transmission vidéo avec très faible latence
Résumé
Applications such as teleoperated driving, remote robot control and telepresence rely on video services to guarantee real-time interaction with a satisfactory quality of experience. Reducing the glass-to-glass delay (G2G), i.e. the time elapsing between the acquisition of a video image and its display on a remote terminal, is essential for these applications. Deep learning-based video image extrapolation has recently been seen as an interesting solution for reducing G2G delay, but reducing latency leads to a loss of quality. The aim of this work is to examine the effectiveness of extrapolation by studying the trade-off between latency and quality.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|