• DocumentCode
    2495668
  • Title

    Compression artifacts reduction for MPEG-2 video utilizing total variation regularization method on CELL processor

  • Author

    Tanaka, M. ; Sakuta, Y. ; Goto, T. ; Hirano, S. ; Sakurai, M.

  • Author_Institution
    Dept. of Comput. Sci. & Eng., Nagoya Inst. of Technol., Nagoya, Japan
  • fYear
    2012
  • fDate
    2-5 Oct. 2012
  • Firstpage
    140
  • Lastpage
    141
  • Abstract
    In this paper, we propose an effective image compression noise reduction system based on the total variation (TV) regularization method. Our system allows the removal of MPEG-2 compression noise such as mosquito noise and blocky noise without losing picture sharpness. This method is particularly useful for HDTV terrestrial broadcasting with insufficient transmission bandwidth for noise-free HDTV pictures. The problem of the TV regularization method application to motion pictures is its high computational time requirement. We solved this problem by using the multi-core CELL processor. The experimental results show that the processing time of our system on the CELL processor is 2.5 ms, which is much lower than the television frame time of 16.7 ms. This shows that our system can be implemented in HDTV receivers.
  • Keywords
    computational complexity; digital video broadcasting; high definition television; image denoising; multiprocessing systems; video coding; HDTV receivers; HDTV terrestrial broadcasting; MPEG-2 video; TV regularization method application; blocky noise; compression artifacts reduction; high computational time requirement; image compression noise reduction system; mosquito noise; multicore CELL processor; noise-free pictures; time 16.7 ms; time 2.5 ms; total variation regularization method utilization; HDTV; Image coding; Image edge detection; Image resolution; Noise; Signal resolution;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Consumer Electronics (GCCE), 2012 IEEE 1st Global Conference on
  • Conference_Location
    Tokyo
  • Print_ISBN
    978-1-4673-1500-5
  • Type

    conf

  • DOI
    10.1109/GCCE.2012.6379558
  • Filename
    6379558