A classification of video games based on game characteristics linked to video coding complexity

Applications used for video streaming of gaming content have seen tremendous growth over the recent years as evident with the increasing popularity of services such as Twitch.tv and YouTubeGaming. Gaming video streaming encoding needs to be performed in real-time and thus has a strict set of encoding constraints. Therefore, many traditional encoding optimization methods such as multiple-pass encoding are not suitable for live gaming video streaming applications. The video quality of streaming services is highly content dependent. While this holds true also for conventional contents, there exist many characteristics of games that do not vary much over time. Therefore, such game-specific information can be exploited to optimize the encoding process. In this paper, we present a classification of games using characteristics such as the type of camera movement, texture details, and static areas of a scene. Using a database of gaming videos from different genres and complexity, we obtain clusters corresponding to the calculated quality values (VMAF). The derived gaming characteristics are then mapped to the quality classes to obtain a decision tree based game classification. We illustrate how the classification can be used for encoding bitrate selection and quality prediction.