Re: IDL virtual reality (was 3D Object IDL )

[george(at)apg.ph.ucl.ac.uk (george Millward)]

Hi There,

Many thanks for the help on this one.  It seems that full virtual
reality is actually very possible.  Rick, thanks for the programs - I
really wasn't expecting custom built applications - but I am glad that
I have got some people excited.

This was a question from Dick:
> (Question: can you really get good pixel-for-pixel alignment across the
> whole display with two projectors? I've never seen it done, but of course
> this will be important for good results.)

I have been working with a 3D video film maker for a while now.  We
have been making a variety of 3D films and the best way to show these
is to use the 2 projector system.  It is bright, clear and can easily
be shown to between 50 and 100 people with no problems.  All you do is
put a polarising filter over each projector.  This ain't fancy - we
have always hired projectors and just taped a filter over the front. 
The 2 filters are set so the polarisation is 1) vertical and 2)
horizontal.  You can then buy polarising glasses.  These are just
cheap plastic glasses with polarising filters which are aligned
horizontal and vertical for each eye.  I am not sure which "lens" in
the glasses is horizontal and which vertical - but obviously this
needs to be set up so that the left eye is getting the light from the
projector with the "left" scene, and likewise the right eye.  OK - I
imagine I am really just stating the obvious here.  Now, the 2 images
from the 2 projectors are aligned on the screen.  But (this is the
answer to Dicks question) the actual accuracy of the alignment is not
as important as he implies.  Ideally, you would want pixel by pixel
accuracy - and the 2 projectors to be colocated etc. - but in such an
ideal situation you would also want the viewer to be located totally
"on axis".  In reality, any slight deviations from this perfection is
sorted out by the brain.  I.E., suppose you are projecting a sphere
into your 3D space ( and lets face it - spheres are pretty popular). 
If some aspect of the system means that the two images are not totally
properly located - then all that happens is that the brain
compensates.  you may have the impression of an object which isn't
perfectly spherical - but not enough to ever give you worries.  Other
practical problems are the fact that the viewers are at different
distances from the screen, and some are off to left/right etc. etc. 
The best effect is when the viewer is located relatively far back and
"on axis".  But all that happens as you move nearer is that the effect
becomes more "radical" - the sphere moves towards you - appears very
much to be hovering in space - and you eventually embarrass yourself
by trying to reach out and touch it !!  But I don't think that pixel
by pixel alignment is totally necessary.  Remember the left eye only
sees what is coming from the left projector
(and likewise for the right).  Even if the 2 projectered images were
misaligned by a centimetre or so - the brain would still recombine the
separate images as a 3D scene.
From our experiences, using twin projectors is by far the best method
for full virtual reality.  You get a bright, full colour, totally 3D
image with no flicker (flicker is a problem with the flicker glasses -
unless you get a very high-end projection system that works at twice
normal refresh rate - Also, flicker glasses are expensive and you have
50 people wired together !!).
I will dig out some literature on the polarising glasses - but I bet
10 minutes on google would give you the answers.
Lastly, our work so far has been video films - these are at
comparatively low-resolution (i.e. standard TV resolutions).  With IDL
driving relatively modest projectors you could easily have a full 1024
* 768 system - and even higher if you can get someone to part with the
money.
The whole system then becomes:  computer with double head graphics
card -> twin projectors -> 2 polarising filters taped over the lenses
-> polarising glasses.  I haven't done it with IDL yet but I reckon
the results will be spectacular!!
I think the best way to set this all up would be to create the full 3D
system and then sit down with the glasses on and "tamper" with all the
free parameters (distance between the "IDL cameras"), i.e. just build
a widget interface and adjust until you get the best results.

OK, I have written too much - time to get on with it.
I will report back with any progress.

Cheers,

George.