#Stereoscopic Images
19 messages in this thread
I am interested in creating a stereoscopic pair of still images of a scene. My
first inclination was to render each of the images at an "eye's distance"
apart. This would be about 2.75" apart if I use myself as a guide. Will this
yield a good result?
Other info:
-The scene is a building interior.
-The furthest object away is 150'.
-There are no objects close to the eye (but I could put stuff up close if it
would make the image more dramatic.
Any other helpful hints would be most welcome.
Bob
Bob,
That will work just fine.
You should also experiment with larger inter-occular distances for big vistas,
this is known as hyper-stereo. While not truly realistic it can be more
artistically pleasing.
Also try experimenting with convergence/toe-in. This will affect where the
object of interest lies in the 3-D window. Parallel cameras work but are not
really correct. It is normal for human eyeballs to accomodate slightly when
focusing, but it is not always necessary in stereoscopy to have the left and
right views converge.
Bruce N. Goren
Bruce:
Thanks for your response. Naturally I have more questions:
1. 'Convergence/toe in' is a new term for me. From context am I correct in
figuring this is describing the distance between the look at points for each
eye?
2. I'm really looking for drama vs. realism so should I spread the eyes and
have them look at a coincidental point in space? Parallel points in space? Or
should I have the look at points farther away than parallel?
3. Is there any benefit to varying from a human FOV?
4. Are there any methods for testing the images without going all the way to
transparencies?
Thanks
Bob
Bob,
1. It's the angle formed with the object at the vertex when the eyes are not
looking parallel.
2. Start with normally eye-spaced parallel cameras, then experiment — there's
no instant formula for dramatic looks other than good photographic sense of
composition and style.
3. As Above.
4. Free viewing stereo pairs is one method, hard trick to master. Try
converting the pairs to a red/blue anaglyph. There's shareware and commercial
programs that can create the combos.
Bruce
Bruce:
Thanks again for the tips. I'll see if I can't come up with something worth
looking at.
Bob
Robert
You might have found a program to do it automatically for you (like Rayman) but
if your going to manually move the camera and want convergence/toe in, the
images will be wrong if you just move the camera. The reason is the quaterion
mathmatics or something that 3DS uses. What I did to get correct convergence
was first place the camera target on what I want the focus point. Then link
everything in the keyframer and rotate it till the camera line is horizontal.
Now render. Then measure the length of the camera line, move it to new angle
and use the tape to get it back at the same distance and render again. The
reason to do this because it the camera and camera target are not on the same
hrizontal plane then moving the camera will tilt the view. This effect is more
pronounced the more you look down or up such as looking down a skyscraper. Hope
I didn't confuse you and that it helps, I know there is a lot out there to auto
this.
—————————–Will
Will:
Yeah, I think you did confuse me a bit. Wouldn't I get the effect I want by
setting up one camera and target and then copying it to a point an eye distance
away? I'm only looking for one still shot so I don't understand the linking
and rotating you mention in the Keyframer.
Bob
Bob,
load oldcity3 and in the top view move the camera above the road
intersection. Now move the camera in an arc and notice in the perspective
window how it tilts, the horizontal moves to a different plane. If you were to
move the camera 2 units in this model the difference would not be that notice
able, but if you move it alot (such as if you want to make the city look like a
toy) it will be very apparent and there will be no 3D effect. If the camera is
at a 45 angle you might be able to get away with just moving it to the new eye
position but it will be technically incorrect and thus maybe harder to
instantly see the 3D image. I'm a perfectionist so I use the method I outlined,
that way I'm sure it's going to be most effective. Note: the reason I rotated
everything in the keyframer is because you set it up how you want it and then
in the KF you can link everything and rotate it, so that the image appears the
same once rotated). Another thing, you probably already know this but if you
want a superimposed red and blue image you can easily use Photoshop or
Photostyler. What I want to know is how do you know what objects will pop out
of the screen? For toe-in images I think its any one that is between the camera
and its target, but I'm not sure. What about for parallel images? I'd like to
know.
Hope this helps,
——————-Will
Will:
/_______________________From your message________________________\\
Now move the camera in an arc and notice in the perspective window how it
tilts, the horizontal moves to a different plane.
__________________________________________________________________
You seem to be describing the situation at the poles – quaternion camera
problem area. The image I am after has the camera nearer the equator and the
target is almost horizontal. The forced quaternion roll is a killer at the
poles but I haven't run into any difficulty elsewhere.
/_______________________From your message________________________\\
Another thing, you probably already know this but if you want a superimposed
red and blue image you can easily use Photoshop or Photostyler.
__________________________________________________________________
The images I'm after will be true color.
Bob
Hi Robert!
Do you know that RayMan has an option to render stereoscopic images? Load your
3DS scene, set the eye distcance, world scale and std. stereo picture distance
and render the scene. The easiest way to render stereo images from your 3DS
files I know.
Michael
Michael:
/_____________From your message_____________\\
Do you know that RayMan has an option to render stereoscopic images?
_______________________________________________
I had no idea, thanks for the heads up. If I were able to get the money for
your program I'd give it a shot. I'm on a seriously limited budget.
Bob
How much is RAYMAN and where do we get it form?
The special price for RayMan is 595 US$ until Jan. 31th (or Feb. 28th, I`m not
sure now). You can get RayMan from Digimation (and other Dealer too but I only
know Digimation). But you should hurry up. Because of the special price (normal
price is 948 US$) and we have a lot of orders so it may be possible to have
some delays if you wait for a long time.
Michael
Is there a 'crippled' version of rayman or something, so I can get a feel for
the speed etc of it?.
I am interested in all info on it b4 purchasing (of course) like does it work
with IPAS 4 etc etc.
We`re thinking about a demo version of RayMan but I don`t believe we`ll have
one for the first release.
If you want more information you can ask Digimation. As you might know RayMan
is not an IPAS but a stand alone render-engine. It reads you complete material
information but doesn`t compute IPAS effects (in the first release) like
WATER.SPX, VAPOR.APX and so on…
Michael
We have a program that will help you. call me at Synthonics, Inc. 818-707-6000
Ray:
Why not tell me what you got right here on ASOFT. Other folks will benefit as
well. Thanks.
Bob
Hi Ray!
Are you an ASOFT lurker??
I talked with Alba yesterday. She mentioned next Wed. as a possible date for
3dmaker??
How did the tape look to you guys? I'd like to try an image or flic at the 752
x 480 x .82 aspect ratio that we use when dumping to tape. Let me know..Also,
for the picture you would use in your PCG&V ad, would we give you a .TGA image
and you turn it into the stereoscopic image, or do we need the 3dmaker program
before we can give you an image.. drop me an email.
Brent Baumann,72202,3562
RTI Animation
can you tell me more about 3dmaker? i am new to 3ds and would like to make 3ds
images. i also need a way to putput my animations if you could email me your
rates . thanks gayle.