Statistical Multiplexing of Transcoded IPTV Streams Based on Content Complexity

摘要

IPTV video services are under development for managed broadband networks, with 'last-mile' delivery across wireless, ADSL or cable access networks. Delivering multiple video streams over a constrained channel often requires bitrate transcoders for bandwidth adaptation. This paper presents a bandwidth allocation scheme based on content complexity to equalize the overall video quality, in effect a form of statistical multiplexing. Complexity metrics serve to estimate the appropriate bandwidth share for each stream, prior to distribution over a wireless access network. These metrics are derived after entropy decoding of the input compressed bit-streams, without the delay resulting from a full decode. Fuzzy logic control serves to adjust the balance between spatial and temporal complexity metrics. The paper examines constant and varying bandwidth scenarios. Experimental results show a significant overall gain in video quality in comparison to a fixed bandwidth allocation.