• DocumentCode
    3149004
  • Title

    Estimating channel-induced distortion in H.264/AVC video without bitstream information

  • Author

    Valenzise, G. ; Magni, S. ; Tagliasacchi, M. ; Tubaro, S.

  • Author_Institution
    Dipt. di Elettron. e Inf., Politec. di Milano, Milan, Italy
  • fYear
    2010
  • fDate
    21-23 June 2010
  • Firstpage
    100
  • Lastpage
    105
  • Abstract
    No-reference video quality monitoring algorithms typically assume the availability of the encoded bitstream in order to assess the quality of the received signal at the decoder side. In some situations this is not possible, e.g. because the bitstream is encrypted or processed by third party decoders. Thus no-reference quality monitoring must be carried out in a blind way, i.e. using only pixel-domain data output by the decoder. In this paper we target this scenario for the specific case of distortion introduced by channel losses. We estimate the missing coding parameters, as well as the channel error pattern, and feed them into a no-reference quality monitoring system which produces accurate estimates of the MSE distortion. The results produced by the proposed method are well correlated (linear correlation coefficient larger than 0.8 over a wide range of packet loss rates) with the distortion computed in full-reference mode.
  • Keywords
    channel estimation; decoding; distortion; video coding; H.264/AVC video; MSE distortion estimation; bitstream information encoding; channel losses; channel-induced distortion estimation; full-reference mode; linear correlation coefficient; no-reference video quality monitoring algorithms; third party decoders; Automatic voltage control; Condition monitoring; Cryptography; Decoding; Distortion; Mean square error methods; PSNR; Video compression; Video on demand; Video sharing;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Quality of Multimedia Experience (QoMEX), 2010 Second International Workshop on
  • Conference_Location
    Trondheim
  • Print_ISBN
    978-1-4244-6959-8
  • Type

    conf

  • DOI
    10.1109/QOMEX.2010.5517880
  • Filename
    5517880