1 / 23

Video Coding on Multi-core Graphics Processors

Video Coding on Multi-core Graphics Processors. Ngai -Man Cheung, Xiaopeng Fan, Oscar C. Au, and Man-Cheung Kung. IEEE SIGNAL PROCESSING MAGAZINE MARCH 2010. Outline. Introduction Motion estimation on GPUs GPU-based ME based on loop unrolling

deanna
Télécharger la présentation

Video Coding on Multi-core Graphics Processors

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Video Coding on Multi-core Graphics Processors Ngai-Man Cheung, Xiaopeng Fan, Oscar C. Au, and Man-Cheung Kung • IEEE SIGNAL PROCESSING MAGAZINE MARCH 2010

  2. Outline • Introduction • Motion estimation on GPUs • GPU-based ME based on loop unrolling • GPU-basedME based on rearranging the encoding order • RD-optimized intra-mode decision on GPUs • Motion compensation on GPUs • GPU-based fast motion estimation • Fast motion estimation • The GPU implementation using tiling • Experiments • My conclusion

  3. Introduction #1 • H.264 • Motion estimation [20][22] • Intra prediction [27] • Motion compensation [10]

  4. Introduction #2 • Challenges • Data dependency • Flow control instructions can degrade the performance • if, switch, do, for, while • Advantages • faster

  5. Motion estimation #1 • GPU-based ME based on loop unrolling [20] • SAD • Metric matching

  6. Motion estimation #2

  7. Motion estimation #3 2~14倍 • Two parts

  8. Motion estimation #4 • GPU-based ME based on rearranging the encoding order • RD-optimized ME 40倍

  9. Intra-mode #1 • RD-optimized intra-mode decision on GPUs [27] • http://www.camdemy.com/media/4371

  10. Intra-mode #2

  11. Intra-mode #3

  12. Intra-mode #4

  13. Motion compensation #1 • Motion compensation on GPUs [10]

  14. Motion compensation #2 4倍

  15. Fast motion estimation #1 • Simplified unsymmetrical multi-hexagon search [42] • Good tradeoff between computational complexity and coding efficiency. • minimizing the Lagrangian cost

  16. Fast motion estimation #2

  17. Fast motion estimation #3 • The GPU implementation using tiling • Neighboring block in another tile is unavailable.

  18. Experiments #1

  19. Experiments #2

  20. Experiments #3

  21. Experiments #4

  22. Experiments #5

  23. Conclusion • Tradeoff between speedup and performance is importance. • There seems to be no study on partitioning the encoding flow between CPUs and GPUs.

More Related