Unbalanced quantized multiple state video coding

Abstract

Multiple state video coding (MSVC) is a multiple descriptionscheme based on frame-wise splitting of the video sequence intotwo or more subsequences. Each subsequence is encoded separatelyto generate descriptions which can be decoded independently. Dueto subsequence splitting, the prediction gain decreases. But sincereconstruction capabilities improve, error resilience of thesystem increases. Our focus is on multiple state video coding withunbalanced quantized descriptions, which is particularlyinteresting for video streaming applications over heterogeneousnetworks where path diversity is used and transmission channelshave varying transmission characteristics. The total bitrate iskept constant, while the subsequences are quantized with differentstepsizes depending on the sequence as well as on the transmissionconditions. Our goal is to figure out under which transmissionconditions unbalanced bitstreams lead to good system performancein terms of the average reconstructed PSNR. Besides, weinvestigate the effects of intra-coding on the error resilience ofthe system and show that the sequence characteristics, and inparticular the degree of motion in the sequence, have an importantimpact on the decoding performance. Finally, we propose adistortion model that is the core of an optimized rate allocationstrategy, which is dependent on the network characteristics andstatus as well as on the video sequence characteristics. © 2006 Hindawi Publishing Corporation. All rights reserved.