Personalized QoE Optimization with Edge-Aided Video Enhancement Services

Video enhancement technologies can not only improve video qualities but also enrich personalized video experiences. With these technologies, the requested videos can be enhanced either on end devices or cloud servers. However, client devices may not have sufficient computing resources or energy, while cloud servers may not satisfy users' short delay requirement. Motivated by the 5G and edge computing technologies, we present veEdge in this paper, an edge-assisted video transmission framework for general-purpose video enhancement services. To balance the trade-off between video enhancement quality and processing delay, we formulate an optimization problem considering video requesting, processing, and playback time simultaneously. To solve the problem in an online fashion, we develop a deep reinforcement learning (DRL) based approach to select the best video enhancement model in favor of various targets, such as bandwidth and quality. The experimental results show that our solutions can significantly improve users' quality of experience (QoE).