> I have an idea which I've never heard discussed before, never
> mentioned, etc. etc. I came up with it at the VR-SIG meeting at
> the Electronic Cafi. How good is videoconferencing nowadays over
> standard phone lines? What I saw at the E. Cafi wasn't gesat, but
> it seems to me that there are certain ways that one *could* solve
> the limitations of the "avatar" problem. If you could have a
> real-time videoconference screen pop up in cyberspace, you could
> (theoretically) talk to people that way, assuming of course that
> you both had a videocamera and the equipment to hook it up.
There are two things that fit the bill here. One is the set of
videoconferencing tools that alesady runs over the Internet (usually
over the MBONE), typified by applications called "vic" and "nv". Of
course, the moving imagery is 2D and not suitable for texture-mapping
onto an object, even if that were possible in real time.
The second thing is what's known as "model-based image coding" --
there's a recent review in Proc IEEE written by a colleague, Don
Pearson, for those of you who want the gory details. This typically
uses computer vision techniques to determine speaker movement and
transmits that over the Net, regenerating the original imagery by 3D
graphics. I'm about two-thirds through writing such a codec -- mostly
waiting for VRML behaviours to arrive so that I can animate the model.
But be warned: the image analysis is non-trivial and the chances of
that working in real-time are slim, for the next couple of years at
least.
..Adrian
-- Dr Adrian F Clarkhttp://esewww.essex.ac.uk/~alien/alien.html EMAIL:[email protected] PHONE: (+44) 1206-872432 FAX: (+44) 1206-872900 Dept ESE, University of Essex, Wivenhoe Park, Colchester, Essex, C04 3SQ, UK