Network assessment for low bitrate video

Abstract

Assessing the impact of network performance on video quality of experience is critical for successful video deployments in enterprises. Typically, network assessments review network LAN and WAN capabilities, QoS mechanisms, and ensure that adequate bandwidth is allocated for video applications. However, most techniques cannot accurately relate network performance to video quality since typically network measurements are poor indicators of video quality. The impact of network degradation on quality depends on many factors (e.g. the type of video frame impacted by the fault, error concealment capabilities of the video codec, application level QoS control mechanisms etc.) and complicates developing a general model of video quality based on network measurements. In this paper, we address the issues associated with developing such a general model by focusing on a subset of video applications with characteristics better suited to network measurement based quality models. Specifically, we focus on low bitrate video conferencing applications with small picture sizes (CIF, QCIF) and low motion levels. In this paper, we study the feasibility of modeling the quality of such applications using network measurements. We report the results of experiments we conducted using the low bitrate configurations of a video phone application running on a PC. For the configurations we considered, our analysis of video call session traces of this application under varying network conditions indicates that network loss measurements estimate video quality reasonably well.