1 / 24

Multi-CPU Video Processing

Multi-CPU Video Processing. Igor Jánoš. Goal of This Project. Decode and process a full-HD video clip using only software resources Dimension – 1920 x 1080 pixels. Motivation. High computational power demands for HD video Latest trend – integrate more CPU cores in one chip

hsayers
Télécharger la présentation

Multi-CPU Video Processing

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. Multi-CPU Video Processing Igor Jánoš

  2. Goal of This Project • Decode and process a full-HD video clip using only software resources • Dimension – 1920 x 1080 pixels

  3. Motivation • High computational power demands for HD video • Latest trend – integrate more CPU cores in one chip • Systems without powerful GPU • GPU is busy with another job

  4. Video Filters • Deinterlace • Noise reduction • Color controls • No dependencies among resulting pixels • Very good for thread-parallel processing

  5. Deinterlace Filter • Produces frames at doubled rate – higher memory demands • Combines odd and even lines from successive frames • Half-pixel shift to reduce comb effect • Reads 2 frames • Writes 2 frames

  6. Previous Frame Current Frame Generated Interlaced Frame Output Frame 1 Output Frame 2

  7. Denoise Filter • Gradual noise reduction • Threshold – reduces temporal noise • Weighted averaging • More difficult to compute • Reads 2 frames • Writes 1 frame

  8. Two successive frames in a stationary scene

  9. Without Denoise Filter

  10. With Denoise Filter

  11. Color Controls Filter • Reads 1 frame • Writes 1 frame • Easy to compute in Y-U-V color space

  12. Color Controls Filter

  13. Basic algorithm idea • “Divide & Conquer” approach • Assign a part of a larger job for each CPU core

  14. Experiment • Use processing filters with live video and offline content • Standard Definition • High Definition

  15. Results

  16. Results

  17. Conclusion • Software-only thread-parallel processing is possible in real-time • Memory performance is the bottleneck • DI performance drop after exceeding memory page size • Latest Quad-Core CPUs from Intel with enlarged L2 cache - 40% faster

  18. Future Work • Different frame buffer structure • Hybrid encoder/decoder (GPU + Multi-core CPU) • GPU – motion estimation & compensation, color space conversion, deblocking • CPU – VLC, transformations, intra predictions

  19. Thank you

More Related