Distributed coding of highly correlated image sequences with motion-compensated temporal wavelets

Abstract

This paper discusses robust coding of visual content for adistributed multimedia system. The system encodes independentlytwo correlated video signals and reconstructs them jointly at acentral decoder. The video signals are captured from a dynamicscene, where each signal is robustly coded by a motion-compensatedHaar wavelet. The efficiency of the decoder is improved by adisparity analysis of the first image pair of the sequences,followed by disparity compensation of the remaining images of onesequence. We investigate how this scene analysis at the decodercan improve the coding efficiency. At the decoder, one videosignal is used as side information to decode efficiently thesecond video signal. Additional bitrate savings can be obtainedwith disparity compensation at the decoder. Further, we addressthe theoretical problem of distributed coding of video signals inthe presence of correlated video side information. We utilize amotion-compensated spatiotemporal transform to decorrelate eachvideo signal. For certain assumptions, the optimalmotion- compensated spatiotemporal transform for video coding withvideo side information at high rates is derived. It is shown thatthe motion-compensated Haar wavelet belongs to this class oftransforms. Given the correlation of the video side information,the theoretical bitrate reduction for the distributed codingscheme is investigated. Interestingly, the efficiency ofmultiview side information is dependent on the level of temporaldecorrelation: for a given correlation SNR of the sideinformation, bitrate savings due to side information aredecreasing with improved temporal decorrelation. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.