A Skill-Based Visual Attention Model for Cloud Gaming

Despite its recent advances and increasing industrial interest, cloud gaming’s high bandwidth usage is still one of major challenges. In this paper, we demonstrate how incorporating visual attention into gaming helps to reduce bitrate without negatively affecting the player’s quality experience. We show that current models, which work well for normal videos, underperform in context videos. Hence, propose our novel model, by developing a skill-based based on dataset. First, it demonstrated players’ maps are correlated with their skill levels can be exploited improve accuracy modeling. Then, fact used cluster maps, according level. A simple yet effective method introduced predict using performance game. Finally, models incorporated video encoder perceptually optimize allocation. Incorporating level model improves saliency 14% respect baseline, 24% competing methods, terms Normalized Scanpath Saliency (NSS). Furthermore, maximum possible amount reduction depends Experimental results 13%, 5%, 15% beginner, intermediate, expert players, respectively.