A comparison of the performance of prediction techniques in curtailing uplink transmission and energy requirements in mobile free-viewpoint video applications

Abstract

The rapid deployment of multimedia services on mobile networks together with the increase in consumer demand for immersive entertainment have paved the way for innovative video representations. Amongst these new applications is free-viewpoint video (FVV), whereby a scene is captured by an array of cameras distributed around a site to allow the user to alter the viewing perspective on demand, creating a three-dimensional (3D) effect. The implementation on mobile infrastructures is however still hindered by intrinsic wireless limitations, such as bandwidth constraints and limited battery power. To this effect, this paper presents a solution that reduces the number of uplink requests performed by the mobile terminal through view prediction techniques. The implementation and performance of four distinct prediction algorithms in anticipating the next viewpoint request by a mobile user in a typical FVV system are compared and contrasted. Additionally, each solution removes the jitter experienced by the user whilst moving from a view pattern to another by allowing some hysterisis in the convergence signal. Thus, this technique enhances the performance of all the algorithms by taking into consideration the fact that the user adapts to the presented views and will react accordingly. Simulation results illustrate that an uplink transmission reduction of up to 96.7% can be achieved in a conventional FVV simulation scenario. Therefore, the application of prediction schemes can drastically reduce the mobile terminal’s power consumption and bandwidth resource requirements on the uplink channel.