• DocumentCode
    3455477
  • Title

    Lagrange multiplier selection for video coding with varying quantization

  • Author

    Aminlou, Alireza ; Yazdanpanah, A. ; Hashemi, Mohammed R.

  • Author_Institution
    Multimedia Process. Lab., Univ. of Tehran, Tehran, Iran
  • fYear
    2011
  • fDate
    8-11 May 2011
  • Abstract
    Lagrange multiplier selection that is widely used in mode decision has a significant effect on the quality performance of video encoding. It has been shown in the literature that the Lagrange multiplier is closely related to the quantization factor. However, changing the Lagrange multiplier based on quantization, as it has been the case in several researches, may not be totally accurate. In this paper, it has been discussed from both analytical and experimental points of view that there is always a constant Lagrange multiplier which provides a better PSNR quality compared to the conventional adaptive Lagrange multiplier. This finding is very useful in applications such as constant bitrate coding or video coding with region of interest. When coding a video with constant bitrate, there is always a constant Lambda value that provides a PSNR quality better than a video encoded at the same bitrate but with an adaptive Lagrange multiplier. The proposed method has also been tested for very low bitrate applications where region of interest coding is used. Simulation results indicate that this approach improves quality by 0.67 dB in average.
  • Keywords
    quantisation (signal); video coding; Lagrange multiplier selection; Lambda value; PSNR quality; constant bitrate coding; quantization factor; region of interest; video coding; Adaptation models; Bit rate; Encoding; PSNR; Quantization; Simulation; Video coding; H.264/AVC standard; Lagrange multiplier selection; rate control; rate-distortion optimization; region of interest;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Electrical and Computer Engineering (CCECE), 2011 24th Canadian Conference on
  • Conference_Location
    Niagara Falls, ON
  • ISSN
    0840-7789
  • Print_ISBN
    978-1-4244-9788-1
  • Electronic_ISBN
    0840-7789
  • Type

    conf

  • DOI
    10.1109/CCECE.2011.6030576
  • Filename
    6030576