A Gaze-based Real-time and Low Complexity No-reference Video Quality Assessment Technique for Video Gaming

Various types of online video gaming services have emerged as video game popularity have increased. Network fluctuations, however, greatly affect the quality of service (QoS) in such gaming services. In addition, video gaming requires a faster response time than general video services. A real-time assessment of quality based on each genre's specific characteristics is another challenge. The objective of this paper is to provide an assessment method based on the user's gaze for the quality of experience (QoE) associated with no-reference video gaming. By exploiting the fact that delay requirements and gaze patterns (e.g., times) differ based on the game genre, the proposed model extracts video features and assesses video quality in real-time. By extracting image features through the human visual system (HVS) approach, we improve the performance of low-performing image features. The proposed model is then trained and verified on popular video gaming datasets (e.g., GamingVideoSET and KUGVD). Through extensive simulation results, we show that certain image features have a higher correlation with the actual mean opinion score (MOS) when HVS is applied. The correlation of some image features is reduced when using the entire frame, but it is not significant, and the calculation time is reduced. Additionally, we demonstrate that the predicted MOS shows a correlation of 0.9 or more with the actual MOS.