Vista: A Semantic Communication Approach for Video Transmission

Recently, video transmission via ultra‐reliable and low‐latency communication (URLLC) is becoming increasingly vital for various multimedia services. However, the burgeoning demand for high‐quality video content places a significant strain on wireless networks, especially considering their limited bandwidth. In response, this paper introduces a novel framework for video transmission, termed VIdeo transmission over Semantic communicaTion Approach (VISTA). VISTA focuses on transmitting key semantic information instead of all video bits, aiming to lower bandwidth usage while maintaining high visual quality. It begins with a semantic segmentation module that separately processes the dynamic and static parts of the video. Subsequently, semantic location graphs (SLGs) are created to capture the semantics and spatial relationships of dynamic objects. Utilizing a joint source–channel coding (JSCC) module, adaptable to varying channel conditions, VISTA transmits these encoded semantic features and SLGs over the wireless channel. At the receiving end, the video is reconstructed using the received semantic features and SLGs, aided by a frame interpolation module. Simulation results show that VISTA surpasses other benchmarks in bandwidth efficiency and channel noise robustness, fulfilling the stringent criteria of URLLC.