2002 Project Plans

This note describes what was accomplished during the third year of the NSF grant (August 2001-July 2002) and a summary of our plans for the next couple of months.

2001-2 Accomplishments

The following list describes the current status on all of the projects planned last year (see 2001 Project Plans for detailed description of each task). The comments about each task reflect our planning when we thought the project would continue.

1. [must] H.263 codec.
   
   A new H.263 decoder is being implemented and integrated into OM. The decoder currently handles intra-coded pictures. Inter-coded motion vectors are currently being debugged. The decoder is efficient. It will be released in the OM 5.3 release planned for September 2002.
   
   
2. [must] 16-bit audio support and audio codecs.
   
   A new 16-bit audio agent has been implemented. It supports stereo 16-bit sampling at rates been 8-44 KHz. This code has been tested in vat although it is not currently available through the user interface. Some testing was done with rat but not a lot. We decided to stop development of this code since the majority of the research community has moved to UCL rat and it did not make sense to continue to replicate their work. Our plan is to switch to that code and provide support for it.
   
   
3. [must] RGB capture abstractions.
   
   This code has been used in several webcasts from Berkeley. It was released to Bob Olson at Argonne for him to test, but it was too unstable. Further work needs to be done before it can be reliably used by others. The biggest problem with the existing code is that the auto-sensing code that tells the software the image size and color model does not work. The code should also be tested on the new boards recently released by the folks who made the VisionRGB board we used - namely, the Datapath folks in England.
   
   
4. [must] Resurrect MediaBoard and AS/1 Services.
   
   This task is only partially completed. We have fixed some of the bugs that kept this code from working, but we still have a problem when initializing a second MediaBoard tool when it joins a session. The AS/1 code was modified and incorporated into the INDIVA prototype, discussed below.
   
   
5. [must] Improved remote video capture application (rvc/rvic).
   
   We have implemented several remote capture tools. The OM project has focussed on the development of the INDIVA software which includes new remote capture services and viewing tools. The AG folks have been developing a replacement for UCL vic using OM.
   
   
6. [must] Upgrade to recent versions of Tcl, Tk, oTcl, and TclCL.
   
   Completed in 5.1.6. We upgraded to Tcl/Tk 8.3.3 and integrated bug fixes from TclCL.
   
   
7. [stretch] Combined audio/video player with synchronization.
   
   The iview tool being developed for INDIVA supports audio and video playback.
   
   
8. [stretch] Memory leak code clean-up.
   
   This problem seemed to be solved in 5.1.6. The biggest problems were in vic which did not destroy objects associated with a session when switching to a new session.
   
   
9. [stretch] Distribute Internet Webcasting System Software.
   
   We decided not to distributed the existing code. After discussions both locally and with potential users elsewhere, we realized the current architecture had too many dependencies on other code (e.g., postgres DBMS, etc.) and was too unstable. We developed a new architecture for the control software middleware, see INDIVA: Distributed Streaming Media and Equipment Control Middleware for more details, and built a prototype that we are deploying in Berkeley for use in our webcasting and collaboration research this coming year.
   
   
10. [stretch] MacOS X Port.
    
    We have completed a port of vat and are almost completed with the vic port. We will produce a binary release of this software in the near future.
    
    

We also completed several projects that were not included in the plan. Some of these tasks resulted from support requests and some resulted from masters student projects. These tasks included:

1. vic improvements.
   
   Numerous requests were made by the AG developers to improve vic. The following improvements and bug fixes were completed. First, we incorporated all known upgrades from the UCL vic code into OM vic, which will be referred to as uclvic and omvic, respectively. Second, the switch session command was fixed so moving between AG rooms does not require terminating and restarting the process. Third, a number of features were added to support placement of thumbnails (e.g., multiple columns, scrollbars, etc.). Fourth, an autoplace mechanism was defined that allowed user-customization of the thumbnail layout. And lastly, a smoothing filter was added to improve the quality of images displayed. The following paper describes the filter and its performance, Implementation of a Real-Time Software-only Image Smoothing Filter for a Block-transform Video Codec.
   
   
2. Ported Nsync to Open Mash.
   
   The Nsync synchronization toolkit was ported to the system. Nsync has a rules and event system for controlling course synchronization between multimedia events. It also incorporates time objects so you can synchronize to absolute or relative time. An efficient implementation was built in Tcl so that code is executed only when necessary. More details about the system are described in the paper Nsync - A Toolkit for Building Interactive Multimedia Presentations. A paper describing the system was presented at ACM Multimedia '98.
   
   
3. RTPtv.
   
   The RTPtv code was significantly enhanced and distributed. More details on this software is available at Production Quality Internet Television. The OM code for handling streaming JPEG was modified so that omvic can play RTPtv streams. Specifically, code was added to handle restart markers, quanitization tables in the stream, and interlaced pictures.
   
   
4. Transcoding Gateway.
   
   A transcoding gateway process that fit into the Director's Console transmission abstraction was implemented. The gateway handles transcoding to alternative formats (e.g., Real Networks), and different sessions and bandwidths. More details are available at tgw: A Webcast Transcoding Gateway.
   
   
5. Titling FX on vic.
   
   A new vic capability is a Special FX engine, that currently is limited to titling addition. The code is part of a research effort at the Computer Science Department. The FX engine processes raw images as they flow from the capture device and add a user-defined, high-quality (Postscript) caption in the bottom part. Current settings are the text contents, font type and size. For more information, please contact José María González.

The port to Solaris was completed, and numerous bugs were fixed.

2002 Plans

This section describes the plans for this coming year. Some of these develops focus primarily on research activities at Berkeley. Others involve finishing and releasing code developed in the last year. Our plan is to do a release in September 2002.

1. H.263 Development.
   
   The decoder will be finished and released. We will also do some work to improve the intra-coded H.263 encoder previously developed by Paul Huang. Currently, that encoder sends every block as an intra-coded block (i.e., essentially sends a JPEG-like stream). We will add a simple extension to only send blocks that changed similar to the H.261 conditional replenishment encoder in vic.
   
   
2. Macintosh Ports.
   
   The vic port will be finished and a binary distribution released.
   
   
3. INDIVA.
   
   INDIVA will be deployed in two locations at Berkeley: 1) the 530 Soda Hall Broadcast Center and 2) the 327 Soda Hall AG-Colab. The goal is to continue the development and testing of this prototype. Several new services will be developed including: 1) a DelcoBox service (i.e., RTPtv), 2) an NCast Telepresenter service, and 3) several camera control services (e.g., Parkervision, Canon VCC-3 and VCC-4, etc.).

   This middleware will be used to build software to automate production of webcasts and AG conferences. A collaboration control process will be implemented for the AG-Colab using Nsync and INDIVA. The goal is to begin the development of software to reduce the overhead and complexity of using the AG. The primary focus will be on automating the startup of an AG conference and the control of window placement, cameras, and specialized streaming services.
   
   

4. Configuration Support.
   
   Design requirements and a prototype will be developed for saving tool configurations for AG rooms. This work is required to automate AG conferences.