1 / 23

Presents IMTC/Wainhouse Research European Forum 2003

Presents IMTC/Wainhouse Research European Forum 2003. Putting H.264 into Good Use: High-Resolution Video Conferencing. Presented by Stephan Wenger TeleSuite Corp. Outline. Where are we now? What is feasible now? A Few Tricks A Real-World System Cross-Codec Rate Control

rossa
Download Presentation

Presents IMTC/Wainhouse Research European Forum 2003

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. Presents IMTC/Wainhouse Research European Forum 2003

  2. Putting H.264 into Good Use:High-Resolution Video Conferencing Presented by Stephan Wenger TeleSuite Corp.

  3. Outline • Where are we now? • What is feasible now? • A Few Tricks • A Real-World System • Cross-Codec Rate Control • MCU-less Continuous Presence • Summary and Outlook IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  4. Where Are We Now? Assumption: High-End, Professional Video-Conferencing • H.263+ or H.263++ (w/ Annex U for background restoration) • 384 kbit/s or 768 kbit/s • At these bit rates, and with cooperative users, few coding artifacts • TDM/ISDN (H.320) or IP (H.323) • CIF (352x288) or CIF-Interlace 2x(352x288) • Capture and rendering typically interlace • 25/30 fps, 50/60 fields per second • Round-trip delay around 400ms + • Round-trip delay for multipoint (w/ MCU) more than one second IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  5. And the Result is… IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  6. What is Feasible Now? Assumptions • Reasonable BOM, small number of DSPs • Full frame rate (25Hz / 30Hz) and progressive scan pictures What is Possible • H.264 video encoding (in a product design) • 640x480 at full frame rate on one high-end DSP • Limit: RAM bandwidth • Bit rate around 384 kbit/s yields acceptable picture quality • Subjective quality significantly better than H.263++ w/ CIF at same bit rate • Objective quality (PSNR) slightly lower • but keep in mind that the pixels are much smaller IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  7. A Few Tricks (Point-to-Point) Identified Problem: Low Quality, and High Delay • Progressive scan hardware throughout (no A/D and D/A losses) • Better quality and good delay characteristics • Interlace throughout also has good delay characteristics • 250ms round-trip delay over IP can be demonstrated • A fixed environment that discourages uncooperative use of the technology • E.g. fixed camera position, fixed background, studio lighting, seating • Improves quality by itself…… and saves cycles of the DSPs which can be spent elsewhere • Background restoration through ERPS is very efficient • Variable, content-dependent capture rate • Not just simple frame skipping • Requires picture segmentation beyond the usual MBs/Slices schemes IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  8. A Few Tricks (Multipoint) Identified Problems: Low Quality, and Very High Delay • Both a result of evil transcoding • Transcoding currently required for all “video mixing” applications • To avoid transcoding • Identical media coding schemes • Create all representations needed in the sending endpoint…… and simulcast them • Ideally suited: IP Multicast…… however, also possible with bridging infrastructure need some innovative changes in the control and mux protocols • Without transcoding, vastly improved user experience • No transcoding artifacts whatsoever • No additional end-to-end delay for multipoint when using IP-Multicast • Very little additional delay when IP Multicast is not available • 10 ms or so for packet/stream duplication IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  9. A Real-World System:2560 Camera/Codec Series • Video resolution 2560x480 • Screen size roughly 5x1 m2 , viewing distance minimal 1.5m • Pixel size 2x2 mm2 • IMAX-style experience, you stop watching a TV screen • Single CCD camera with custom, wide-angle lens • Progressive scan, programmable frame rate, CameraLink digital interface, low noise, operation point adjusted to studio light • Multiple DSPs, distributed codec • Codec processes four sub-images of 640x480 • Cross-codec rate control shifts bit rate to where it is needed • Each encoder can generate 640x480…… or 640x240, 320x240, 320x120, 160x120, 160x60 simultaneously • Carrier-grade, compact-PCI hardware architecture • Not cheap, but reliable • Connectivity 1.5 Mbit./s minimum, 3MBit/s typical IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  10. MVA Hardware IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  11. Fixed Environment IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  12. Cross-Codec Rate Control (1/4) Subjects are seated in discussion Data load impact per screen 25% 25% 25% 25% IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  13. Cross-Codec Rate Control (2/4) One participant stands Data load impact per screen 22% 34% 22% 22% IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  14. Cross-Codec Rate Control (3/4) Participant leaves the room Data load impact per screen 40% 20% 20% 20% IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  15. Cross-Codec Rate Control (4/4) Subjects continue their discussion Data load impact per screen 25% 25% 25% 25% IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  16. MCU-less Continuous Presence Effective meeting management with multi-point conferencing Point-to-point meeting IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  17. MCU-less Continuous Presence Effective meeting management with multi-point conferencing Multi-point meeting (3 locations) IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  18. MCU-less Continuous Presence Effective meeting management with multi-point conferencing Multi-point meeting (4 locations, only two camera capture fields used) IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  19. MCU-less Continuous Presence Effective meeting management with multi-point conferencing Multi-point meeting (6 locations) IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  20. And the Result is… TeleSuiteNetwork 50 – 120 hrs/month vs. Traditional “talking heads” utilization 0-5 hrs/month IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  21. Summary: Why is H.264 Essential for TeleSuite? • High compression makes bigger picture formats possible • MPEG-2 quality at 1/3 of the bit rate • ERPS allows for efficient, standard-compliant background restore • No visible “trailing artifacts” • This is useful primarily in a controlled environment • ERPS and FMO allow for good error resilience • Helpful for IP networks and especially IP Multicast • Arbitrary picture sizes commonly implemented • In contrast to H.263+ or MPEG-2 • H.264 is by far less complex than generally assumed • One document, few conformance points help reduce implementation complexity • Very DSP-friendly • Modern technology, marketing argument IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  22. Outlook • Even higher resolution • 1024 x 768 per field, 4096 x 768 per screen • 21 DSPs today, 13 DSPs of the 2004 generation • Target bit rate between 500kbit/s and 2 Mbit/s per field is realistic • At this resolution, the pixel structure is invisible • 1.3x1.3 mm2 pixels at 1.5m viewing distance • Provides unmatched, “beyond TV” experience • Matched audio performance • Surround-sound audio with appropriate echo canceling • Audio-zoning (especially difficult w/ more complex continuous presence scenarios) • Higher frame rates? • Does it make sense to go beyond 30 fps progressive? IMTC/Wainhouse Research European Forum – May 2003 – Geneva, Switzerland

  23. Thank You Stephan WengerTeleSuite Corp. stewe@telesuite.com

More Related