Towards Scalable 360-Degree Video Streaming: A Method with an Adaptive Selection Enhancement Mechanism

To address the challenges of bandwidth limitations and the decline in quality of experience (QoE) during the transmission of 360-degree videos, this paper proposes a video streaming framework based on scalable video voding (SVC) with an adaptive selection enhancement mechanism (ASEM). The framework formulates two optimization models: a quality decision model (QD) for current 360-degree video and an enhancement decision model (ED) for unplayed video in the buffer. To solve these optimization problems efficiently, deep reinforcement learning (DRL) algorithms are developed. In addition, a predictive execution strategy is introduced to determine when to execute enhancement decisions, thereby improving the quality of buffered unplayed video while minimizing interference with quality levels decision of future video. Extensive experiments demonstrate that the proposed method consistently outperforms baseline approaches across multiple QoE metrics.