Advancing Temporal and Quality Adaptation in Video Streaming with AV1

Abstract

Traditional Adaptive BitRate (ABR) streaming faces the challenge of providing a smooth experience under highly variable network conditions. Many forms of quality adaptation have been proposed, mainly exploring quality or temporal adaptation. The emerging AV1 codec not only advances video encoding and transcoding but also presents opportunities for more flexible quality adaptation. In this thesis, our objective is to push the limit of quality and temporal adaptation with AV1 frame and segment structure. We propose Temporal Adaptive Streaming over QUIC (TASQ) technique and S-frame Adaptive Bitrate (SABR) streaming for temporal and quality adaptation. Our evaluation results show that both TASQ and SABR can significantly improve the smoothness of the streaming playback while conserving bandwidth and improving the Quality- of-Experience (QoE). The only trade-off is short stutters. To smooth out the stutter, we proposed the IFNet-based Frame Filling (IFF) technique. Collectively, TASQ, SABR, and IFF offer a complete quality and temporal adaptation solution.