Fluency in Real-Time Video Streaming by Learning Human Perceptive Traits to Reveal the Expected Section in Outstanding Quality
概要
Currently, the quality of digital media and the quantity of contents are both increasing rapidly. For instance, watching e-sport competitions often suffers from unstable bandwidth, which causes the video to stutter or have a low resolution. In this situation, users will have a negative experience. Many situations can cause problems of congestion in real-time applications or 3D displays. To solve this kind of problem, we attempt to determine an inverse solution according to the path. This project adopts a reverse operation that reduces necessary data but maintains the same quality perception of user experience by utilizing the characteristics of the human vision and brain. To explore our approach, we develop a prototype that changes the resolution of the image according to a user’s habit and shows the part in focus clearly while leaving the resolution of the background lower. To select optimized sub-image in pictures with higher quality and achieve a lower transmission requirement, full quality is reduced. This will allow the user experience smoother streaming when there is congestion or unstable situations. Then, we conduct a preliminary user study to investigate some future directions and explore some potential flaws.