Edge-Assisted Virtual Viewpoint Generation for Immersive Light Field

Light field (LF), which describes the light rays that emanate at each point in a scene, can be used as a six-degrees-of-freedom (6DOF) immersive media. Similar to the traditional multiview video, LF is also captured by an array of cameras, leading to a large data volume that needs to be streamed from a server to users. When a user wishes to watch the scene from a viewpoint that no camera has captured directly, a virtual viewpoint must be rendered in real time from the directly captured viewpoints. This places high requirements on both the computing and caching capabilities of the infrastructure. Edge computing (EC), which brings computation resources closer to users, can be a promising enabler for real-time LF viewpoint rendering. In this article, we present a novel EC-assisted mobile LF delivery framework that is able to cache parts of LF viewpoints in advance and render the requested virtual viewpoints on demand at the edge node or user's device. Numerical results demonstrate that the proposed framework can reduce the average service response latency by 45% and the energy consumption of user equipment by 60% at the cost of 55% additional caching consumption of edge nodes.

Automated summary

The Light field (LF) is a technique that describes the light rays emitted at each point in a scene, and it can be used for six-degrees-of-freedom (6DOF) immersive media. Similar to multiview video, LF is captured by an array of cameras, resulting in a large data volume that needs to be streamed to users. Rendering virtual viewpoints in real-time from the directly captured viewpoints places high demands on computing and caching capabilities. Edge computing (EC) brings computation resources closer to users and can enable real-time LF viewpoint rendering.