Video Quality Monitoring over IP Networks

Accurate measurement of network characteristics is critical to ensure that networks are operating optimally as designed and configured. Present network measurement metrics, such as packet loss ratio and average latency, were designed to assess the impact of network impairments on data applications. However, as IP networks are increasingly used to carry multimedia traffic such as voice-over-IP and video-over-IP, which are much more sensitive to errors than file transfer, e-mail or web traffic, these data-centric network measurement metrics have become insufficient. — For example, highly-compressed MPEG video traffic exhibits a strong content correlation from one packet to the next packet, and thus, the loss of one packet may render a large number of perfectly received packets useless. One study (see [1]) shows that for a given MPEG-compressed stream, a packet loss rate as low as 3% may be translated into a video frame error rate as high as 30%. Therefore, the packet loss rate does not accurately reflect the severity of a network impairment's effect on video quality. — Furthermore, as video carriers increasingly use Forward Error Correction (FEC) techniques to recover packet loss, measurement metrics designed to optimally configure the FEC operating parameters to match the network condition become highly desirable. In this paper, we review a set of traditional network measurement metrics that are commonly used for data applications and discuss the deficiencies of these metrics for providing an accurate picture of video quality. We then discuss how video quality metrics can be derived from network performance metrics using simple equations and propose video-centric network metrics to compute such quality metrics. We also show how these video-centric metrics can be used to derive a probabilistic loss model which in turn can be used to evaluate the performance of a FEC algorithm.

Published

2005-11

Content type

Original Research

DOI

10.5594/M00409

ISBN

978-1-61482-942-3