• DocumentCode
    2632691
  • Title

    Adaptive Golomb Codes For Level Binarization In The H.264/AVC FRExt Lossless Mode

  • Author

    Bardone, Davide ; Carotti, Elias S G ; De Martin, Juan Carlos

  • Author_Institution
    Dipt. di Autom. e Inf., Politec. di Torino, Turin
  • fYear
    2008
  • fDate
    16-19 Dec. 2008
  • Firstpage
    287
  • Lastpage
    291
  • Abstract
    Fidelity Range Extensions (FRExt) is an H.264/AVC amendment which provides enhanced coding tools and the possibility to perform high resolution and lossless video encoding. However, most of the efforts for lossless coding in the H.264/AVC framework have been concentrated on improving the prediction step while leaving the entropy coder; CABAC, originally designed for lossy coding, unaltered. However, if transformation and quantization of the corresponding coefficients are not performed, as is the case of the lossless coding mode for FRExt, CABAC becomes sub-optimal. In this paper we show how considerable improvements in compression ratios can be achieved with simple modifications of the CABAC engine. The proposed technique was tested on a set of 4:4:4 test sequences, achieving gains of up to 12.80% with respect to the original unmodified H.264/AVC algorithm.
  • Keywords
    adaptive codes; data compression; quantisation (signal); video coding; CABAC; H.264/AVC FRExt lossless mode; adaptive Golomb codes; coding tools; compression ratios; entropy coder; fidelity range extensions; high resolution lossless video encoding; level binarization; quantization; transformation; Automatic voltage control; Bit rate; Engines; Entropy; High definition video; IEC standards; ISO standards; Quantization; Testing; Video coding; Lossless; Video Coding;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signal Processing and Information Technology, 2008. ISSPIT 2008. IEEE International Symposium on
  • Conference_Location
    Sarajevo
  • Print_ISBN
    978-1-4244-3554-8
  • Electronic_ISBN
    978-1-4244-3555-5
  • Type

    conf

  • DOI
    10.1109/ISSPIT.2008.4775706
  • Filename
    4775706