Video traffic modeling by exploiting inter-frame correlation coefficients

Abstract

The explosive growth of multimedia applications renders the efficiency network resource allocation a problem of major importance. The burstiness of video traffic, in particular, calls for innovative ideas that will help prevent significant packet losses that could cause bad Quality of Service (QoS) and Quality of Experience (QoE) to users. In this work, we analyze the traffic characteristics of H.264 encoded video and we propose and evaluate the accuracy of a H.264 video traffic model that predicts the size of B-frames. B-frame prediction can be used in order to reduce bandwidth requirements and smoothen the encoded video stream, by selective B-frame dropping, when the model predicts larger B-frame traffic than the network can handle. We implement three prediction models and after comparing them against each other we conclude which model offers the highest accuracy. It will be shown that our approach clearly outperforms another model recently proposed in the literature, and provides a highly accurate prediction for a wide variety of video traces with different GOP patterns and different qualities. We also show that the third implemented model, which in a previous work in the literature was shown to underperform for MPEG-4 traffic, also produces highly accurate results for H.264 traces.