VCIP 2000 Paper, Simulation Results, and Software

Stephan Wenger
stewe@cs.tu-berlin.de

 

Here you can find:

-          pdf version of the VCIP 2000 paper entitled A High Level Syntax for H.26L: First Results 44 KB

-          Reconstructed pictures of Experiment 2 in MPEG format (see the paper for a description). WARNING: 15.1 MB.

-          Reconstructed pictures of Experiment 3 in MPEG format (see the paper for a description). WARNING: 12.1 MB

 

TML-1 now is now outdated as the development of the new H.26L Recommendation continues rapidly. Therefore, I decided not to make publicly available the codec that was used to perform the simulations. If you are interested in contributing to the development of H.26L in general or in its error resilience features in particular, you might want to do so based on the newest accepted test model and/or on the most up-to-date simulation software. This software and all the standardization documentation are available on the Q.15 ftp site at ftp://standard.pictel.com/video-site. On this site, there are directories reflecting the last couple of meetings. Go to the latest meeting and download the document with the highest document number, which contains the meeting report. From here, you should be able to find the most up to date software, and also be able to get a history of the development of TML, which might be helpful to avoid hunting down trails that are known to be dead ends. To attend meetings, you need to contact the Rapporteur of Q.15, Dr. Gary Sullivan, at garysull@microsoft.com. Until now, serious researchers from industry and academia had no problems in getting a meeting invitation for the majority of the meetings, regardless of their organization being ITU member or not.

 

The reconstructed pictures of the experiments were MPEG coded in order to reduce the file size. MPEG is a lossy encoding method and, therefore, introduces new artifacts, which make quality assessment a bit more tricky. The reconstructed pictures were coded using a bitrate of 1.5 MBit/s and the resulting PSNR from this coding process is higher then 42 dB for all pictures, and higher then 45 dB for most. Hence, the unavoidable artifacts are insignificant compared to those introduced by the original coding algorithm and the reconstruction of the damaged bitstream. If anyone still believes that he/she really needs the uncompressed data, then I will gladly burn two CDs with AVI files and ship them by snail mail. Both your and my network administrator would otherwise probably cry learning about ftp downloads of 500+ MB files.

 

Stephan Wenger, May 17, 2000