Foveated Video Streaming for Cloud Gaming

Research output: ThesisMaster's thesis

Abstract

Video gaming is generally a computationally intensive application and to provide a pleasant user experience specialized hardware like Graphic Processing Units may be required. Computational resources and power consumption are constraints which limit visually complex gaming on, for example, laptops, tablets and smart phones. Cloud gaming may be a possible approach towards providing a pleasant gaming experience on thin clients which have limited computational and energy resources. In a cloud gaming architecture, the game-play video is rendered and encoded in the cloud and streamed to a client where it is displayed. User inputs are captured at the client and streamed back to the server, where they are relayed to the game. High quality of experience requires the streamed video to be of high visual quality which translates to substantial downstream bandwidth requirements. The visual perception of the human eye is non-uniform, being maximum along the optical axis of the eye and dropping off rapidly away from it. This phenomenon, called foveation, makes the practice of encoding all areas of a video frame with the same resolution wasteful.

In this thesis, foveated video streaming from a cloud gaming server to a cloud gaming client is investigated. A prototype cloud gaming system with foveated video streaming is implemented. The cloud gaming server of the prototype is configured to encode gameplay video in a foveated fashion based on gaze location data provided by the cloud gaming client. The effect of foveated encoding on the output bitrate of the streamed video is investigated. Measurements are performed using games from various genres and with different player points of view to explore changes in video bitrate with different parameters of foveation. Latencies involved in foveated video streaming for cloud gaming, including latency of the eye tracker used in the thesis, are also briefly discussed.
Original languageEnglish
QualificationMaster's degree
Awarding Institution
  • Aalto University
Publisher
Publication statusPublished - 2017
MoE publication typeG2 Master's thesis, polytechnic Master's thesis

Keywords

  • foveated video streaming
  • cloud gaming
  • real time foveated encoding
  • gaze tracking applications

Fingerprint Dive into the research topics of 'Foveated Video Streaming for Cloud Gaming'. Together they form a unique fingerprint.

Cite this