The exchange of information with low to zero latency is of paramount importance for many emerging applications, such as remote driving, drone control or remote surgery. However, video coding and network transmission times are incompressible.
Linear Video Coding (LVC) is an emerging framework for wireless video transmission to multiple users. Relying solely on linear operators, LVC provides each receiver with video quality proportional to the quality of its channel. However, LVC suffers from significant coding latency.
The Zero-Latency Linear Video Coding (ZL-LVC) project aims to propose an end-to-end video coding and transmission system with low to zero latency. The aim is to optimize the bitrate-latency-distortion trade-off for traditional and LVC systems. The latter can benefit from the integration of LVC and neural networks in the learning phase.
We therefore face three major scientific challenges:
- Study the encoder design choices best suited to latency reduction ;
- Propose efficient latency reduction schemes for conventional and LVC codecs;
- Propose a hybrid video transmission system (conventional + LVC) with arbitrarily low latency.
We will evaluate the potential of ZL-LVC by means of a demonstrator with software-defined radio platforms.
The ZL-LVC project has been funded by the ANR under the 2020 call for projects. The partnership is led by prof. Marco Cagnazzo (LTCI/Telecom-Paris) and comprises four partners: in addition to LTCI, L2S/CentraleSupélec, IEMN/UPHF, and Ektacom.