>I was demonstrating this yesterday at System95 in Germany...... 3D hologram=
s
>which stick out of the monitor, and you can look at them from all direction=
s.
>
>Ok, you need special glasses to see it, but it really works...
All directions, huh? Like from behind the monitor? Anyhow, if you needed=
special glasses I don't think it was a hologram. Probably some kind of=
stereoscopic display. I guess it could have been a holographic stereogram=
(that's where holography is used to encode and display the stereo pairs,=
but they're still just stereo). Does this matter? Maybe not. But as an=
holographer I get irritated when people misuse the term. Sometimes it seems=
like you can almost guarantee that if something says "hologram" on it it=
isn't a hologram!
Don't get me wrong: I love stereo. But it aint the same thing as holography.=
And holography has limits too. But different limits.
The term "3D" is a funny old beast. The stereo-photography crowd complains=
that computer graphics "3D" isn't "real 3D". And I'd have to counter that=
their stereo isn't "real 3D" either. And then there's SIRDS....
Thinking about displays...
Recently Alex Okita <UB@notes.UB.com> wrote:
>What would the perfect most ultimate VRML browser do? Would it let the user=
=20
>control LOD? what kind of parameters would it allow for? this is all w/o=20
>concern for the computer, say you had a CRAY j90 supercomputer or a Daystar=
=20
>MP6... so speed didn't matter... What would the Browser let you control?=
would=20
>there be a stereoscopic option for a VRhsadset? if you touched an object=
would=20
>it let you see its code? i.e. Oh, look her is a ball [touch] a small window=
=20
>appears "sphere{}" yea great... VoiceCommands? RayTraceing? Raydiosity?=
what=20
>kind of network capabilities? Im just trying to get a feel of what people=
would=20
>like to see out of a viewer. both rsalistically and fantastically.
Ok. Well if "speed didn't matter" I'd say to match what I see in the rsal=
world with real eyes it'll have to simultaneously and correctly stimulate=
all of the depth cues. What you see depends in large part on how you see=
it, so this hypothetical browser's gonna have to keep up with a pretty=
slick hypothetical display device. So I'd like it to sense the=
accommodation and convergence of my eyes at frame rate and compute=
view-pairs with the correct viewpoints, distortions, and degrees of defocus=
so that when directly written onto my estinas I see the world correctly in=
"3D". [It has to be direct estinal writing otherwise my eye's will be=
focused on some flat external display and the muscular cue will be wrong=
even if the images are defocused correctly.] And I want it to fill me field=
of view. With one arc-minute resolution at field center. Full color of=
course, with a dynamic range from starlight to noon-day sun. It'll need to=
know the shape and color of my (virtual) nose, so when I cross my eyes I'll=
see it properly! Glints will require that it models polarization effects=
off surfaces properly. Oh, I nearly forgot iridescence. And of course it'll=
need to update fast enough to be flicker free at any brightness level and=
to track my eye's saccades. Then there's cyclotorsion (the eyeballs rotate=
around the line of sight) which I only just learned about and can't pretend=
to understand yet. Somehow I think speed will always matter!
Adjusting my pedant's hat for a moment, I guess what I'm saying is it's=
futile to attempt to compute a copy of the real world so accurately that I=
can't see it's fake. That visual Turing test will always be failed, except=
maybe for the most minimal of scenes like a view of the inside of a=
recently cleaned integrating sphere. Give me even as simple a scene as one=
Lego(R) brick in a dark dust-free room lit by a single shaft of sunlight:=
show it on any display ever built, and I'll still be able to tell I'm not=
seeing the real thing. So however good browsers may get, the VR experience=
will be no more than a convincing alternative to rsality, not a totally=
convincing replacement. And the problem's in the display, not just the=
compute power or model complexity.
None of this means I dislike the programme of visual rsalism Sutherland=
started us all on some thirty years ago. But until someone has practically=
infinite compute power and (I'm guessing) direct input into my brain/mind=
(not just the neurons connected to the optic nerve, but also those for the=
inner ear, tongue, nose, skin, ...) it won't fool me.
Actually what I'd most like in a browser short-term is automatic dirt and=
damage. One of the things that makes the physical world visually=
interesting is that it's imperfect. I like B-movies because they can't=
afford to cover this up. Give me a browser which understands to throw in a=
few specks of dirt. Modulate those uniform colors and flat surfaces. Maybe=
even just adding some noise to the frame buffer would help. Then give=
preference to adding anything the artists can describe: they know how to=
make it worth having a good browser.
Ok. Back to the DEF debate.
Stephen J Hart fax +1 (714) 348-8665
Director of R&D e-mail sjhart@voxel.com
WWW URL http:/www.voxel.com/
VOXEL, 26081 Merit Circle, #117, The opinions and "facts" expressed
Laguna Hills, CA 92653-7017, USA herein are not necessarily correct