• DocumentCode
    3411919
  • Title

    Highly efficient forward two-dimensional DCT module architecture for H.264/SVC

  • Author

    Husemann, Ronaldo ; Majolo, Mariano ; Roesler, Valter ; De Lima, Jose Valdeni ; Susin, Altamiro Amadeu

  • Author_Institution
    Electr. Eng. Dept. (DELET), UFRGS, Porto Alegre, Brazil
  • fYear
    2010
  • fDate
    8-11 June 2010
  • Firstpage
    1
  • Lastpage
    7
  • Abstract
    The emerging H.264 SVC (Scalable Video Coding) standard specifies an encoder solution responsible for generating a multi-layer stream, which provides extra flexibility for modern multimedia applications. The increased data dependency among different layers demands a significant overall encoder performance. Aiming the use of an SVC solution in real-time applications we propose an optimized hardware implementation of the computational forward two-dimensional discrete cosine transform module. The proposed solution introduces a fast pipelined architecture specially designed to explore hardware parallelism and to surmount memory access delays, processing efficiently up to eight pixel samples in a clock cycle. Practical results confirm the proposal as a high efficient solution able to speedup SVC encoder performance with reduced impacts in complexity.
  • Keywords
    data compression; discrete cosine transforms; multimedia systems; video coding; H.264 SVC; clock cycle; data dependency; encoder performance; memory access delays; multilayer stream; multimedia applications; optimized hardware implementation; real-time applications; scalable video coding; two-dimensional DCT module architecture; two-dimensional discrete cosine transform module; Clocks; Complexity theory; Discrete cosine transforms; Hardware; Pixel; Static VAr compensators; Discrete Cosine Transform; video encoder and H.264 SVC standard;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Rapid System Prototyping (RSP), 2010 21st IEEE International Symposium on
  • Conference_Location
    Fairfax, VA
  • Print_ISBN
    978-1-4244-7073-0
  • Electronic_ISBN
    978-1-4244-7072-3
  • Type

    conf

  • DOI
    10.1109/RSP.2010.5656330
  • Filename
    5656330