3rd International ICST Conference on Mobile Multimedia Communications

Research Article

Joint Error Concealment and Error Recovery for Consecutive Frame Losses under the Unbalanced Multiple Description Coding Architecture

Download593 downloads
  • @INPROCEEDINGS{10.4108/ICST.MOBIMEDIA2007.1669,
        author={Feng Huang and Lifeng Sun and Bin Li and Yuzhuo Zhong},
        title={Joint Error Concealment and Error Recovery for Consecutive Frame Losses under the Unbalanced Multiple Description Coding Architecture},
        proceedings={3rd International ICST Conference on Mobile Multimedia Communications},
        proceedings_a={MOBIMEDIA},
        year={2010},
        month={5},
        keywords={Unbalanced Multiple Description Coding Error Concealment Error Recovery Error Propagation.},
        doi={10.4108/ICST.MOBIMEDIA2007.1669}
    }
    
  • Feng Huang
    Lifeng Sun
    Bin Li
    Yuzhuo Zhong
    Year: 2010
    Joint Error Concealment and Error Recovery for Consecutive Frame Losses under the Unbalanced Multiple Description Coding Architecture
    MOBIMEDIA
    ICST
    DOI: 10.4108/ICST.MOBIMEDIA2007.1669
Feng Huang1,*, Lifeng Sun1,*, Bin Li1,*, Yuzhuo Zhong1,*
  • 1: Department of Computer Science and Technology, Tsinghua University FIT 1-512, Tsinghua University Beijing 100084, China +86 1062786910
*Contact email: hfbee@263.net, sunlf@mail.tsinghua.edu.cn, libin98@mails.tsinghua.edu.cn, zyzdcs@mail.tsinghua.edu.cn

Abstract

Real-time video transmission over error-prone wireless networks often experiences consecutive frame losses due to either temporary link outages or traffic congestion. Although error concealment (EC) techniques have been extensively studied, they usually cannot handle the problem. Thus, we envision using EC under the unbalanced multiple description coding (UMDC) architecture. UMDC has almost no coding delay and can produce two descriptions at any bit-rates adaptive to different path bandwidths. In this paper, we propose an iterative EC algorithm able to adaptively exploit both the high-resolution (HR) and low-resolution (LR) information via multi-hypothesis weights. It is applied to both lost HR frames and following undecoded ones. Considering error propagation, we design an interframe error recovery (ER) algorithm for the undecoded HR frames. It iteratively uses multi-frame recovery principle to frame-by-frame reduce error drift in the HR stream with respect to the intermediate information from EC. The joint design of EC and ER can be applied to most UMD approaches. Extensive experiments have been carried out under different conditions. The proposed EC technique exhibits high PSNR gains versus the usual ones under the UMDC architecture and the classical ones without the support of UMDC. The proposed ER technique is efficient in reducing error drift. In conclusion, joint EC and ER can provide satisfactory performance on both PSNR and visual quality.