RGI: Bit-rate modeling of a novel real-time picture-importance scheme for robust video over the LTE downlink

This paper presents a study on modeling a real-time picture-importance scheme, which is a dilemma problem between high accuracy and real-time applicability. A novel scheme called real-time group-of-pictures-averaged index (RGI) is proposed with two objectives: (1) providing a higher accuracy than the conventional real-time schemes such as I/P and EPL, and (2) still meeting the desired real-time constraint. The higher accuracy is achieved based on a numerical modeling of the effect of target streaming bit-rate change on the priority ordering of member pictures in the same group-of-pictures (GOP), and the fitted results to this effect can thus provide an empirical law for a real-time interpolated prediction of picture importance in a video stream of any target bit-rate. Our results show that the superiority of RGI over the I/P and EPL schemes through a simulation validation of real-time video delivery over the LTE downlink, and the performance gains of video quality are substantial under different congestion levels for various bit-rates and moving speeds of user equipments.