Chris Huff <chr### [at] maccom> wrote:

> In article <3a968e3f@news.povray.org>, Geoff Wedig 
> <wed### [at] darwinepbicwruedu> wrote:

>> I've never done it, but it should be possible.  How does one go about 
>> making such a beast?  Other than transforming the camera slightly one 
>> way or t'other, I mean

> Well, I suppose you could put the entire scene into a union and 
> transform it instead...why don't you want to transform the camera?

No, I can transform the camera, I meant "Other than transforming the camera,
what needs done?"  I assume I'd post process them into a side by side image,
but I wasn't sure that I didn't have to transform the camera in a certain
way, or transform the look at as well, or whatnot.  If it's simply
transforming the camera 4 inches to the left (about the distance eyes are
apart, right?) then it's easy.

Geoff

Post a reply to this message

In article <3a96ad70@news.povray.org>, Geoff Wedig 
<wed### [at] darwinepbicwruedu> wrote:

> No, I can transform the camera, I meant "Other than transforming the 
> camera, what needs done?"  I assume I'd post process them into a side 
> by side image, but I wasn't sure that I didn't have to transform the 
> camera in a certain way, or transform the look at as well, or 
> whatnot.  If it's simply transforming the camera 4 inches to the left 
> (about the distance eyes are apart, right?) then it's easy.

You should rotate the cameras around the look_at point so they are about 
3 inches apart (adjust this for the desired effect). Figuring out the 
exact angles to use would require some fairly simple trigonometry...but 
the easiest way would probably just be to move the location of the 
cameras (leave the look_at point alone, so each camera targets the right 
spot). This will work best if you move the location along an axis 
perpendicular to the direction the camera is facing. (hint: look up the 
vcross() function)

-- 
Christopher James Huff
Personal: chr### [at] maccom, http://homepage.mac.com/chrishuff/
TAG: chr### [at] tagpovrayorg, http://tag.povray.org/

<><

Post a reply to this message

The shift will be smaller than you might at first think . . .
I believe that in general . . . shoot for 1 unit of displacement for every
30 units of distance to the target.  If your camera is 30 from the wall,
shift the camera 1 unit to the right.
  Then paste the results side by side.  These are often difficult to view
on a monitor (crossed-eyes method) and the wider the pic, the more
difficult it will be for folks to see.  Any shift > 3 inches on your monitor
will be difficult or impossible for most folks to see.  Prints offer you a
little more latitude.
  I would just simulate your scene and set up a few simple objects at those
distances and experiment.  It's a great effect if you can get it right . . .
and, as I said, the most common mistake . . . even with a camera, is to
shift too far.  The brain (where depth perception lives) can't make sense
of the data.
--
Y

Post a reply to this message

In article <3a96ebbf@news.povray.org>, "yooper" <Out### [at] huntelnet> 
wrote:

>   The shift will be smaller than you might at first think . . .
> I believe that in general . . . shoot for 1 unit of displacement for 
> every 30 units of distance to the target.  If your camera is 30 from 
> the wall, shift the camera 1 unit to the right.

Where does this "rule of thumb" come from? Your eyes don't get farther 
apart when looking at distant objects, they stay the same distance apart 
and rotate to be closer to parallel. And the distance of the separation 
should be fairly close to what it would be if you were standing in the 
scene, or the scale won't look right.

-- 
Christopher James Huff
Personal: chr### [at] maccom, http://homepage.mac.com/chrishuff/
TAG: chr### [at] tagpovrayorg, http://tag.povray.org/

<><

Post a reply to this message

"Chris Huff" <chr### [at] maccom> wrote in message
news:chrishuff-B43B9F.10013925022001@news.povray.org...
> In article <3a96ebbf@news.povray.org>, "yooper" <Out### [at] huntelnet>
> wrote:
>
> >  shoot for 1 unit of displacement for
> > every 30 units of distance to the target.  If your camera is 30 from
> > the wall, shift the camera 1 unit to the right.
>
> Where does this "rule of thumb" come from? Your eyes don't get farther
> apart when looking at distant objects, they stay the same distance apart
> and rotate to be closer to parallel. And the distance of the separation
> should be fairly close to what it would be if you were standing in the
> scene, or the scale won't look right.

I had read somewhere that 1 unit separation for a focal point 300 units away
was a good idea for the average scene.  Of course the best way would have to
be to use units as a real world measurement, say inches, and go from there.
Obviously still needs to consider field of view (angle or direction in
camera).
I agree with Chris here that following actual human eye characteristics must
be the way to go, however maybe the reasoning not to do so has to do with 3D
photography in some way.  I guess most any camera was/is setup to look
parallel because of perspective changes from the parallax eyes usually have
would cause some kind of non-alignment trouble.
It's not something I've truly learned anything about so I'm just chattering
mindlessly here.
The stereo pairs I posted just now here at p.b.i. have more than a 5 unit
shift with focal point some 600 units away, the camera being rotated around
that.  Scale based on that would make the scene just room-sized and not the
large open spaces it is supposed to be.  25 feet (7.6 meters) to the middle
of the image.

Bob H.

Post a reply to this message

> > Chris:
> > Where does this "rule of thumb" come from? Your eyes don't get farther
> > apart when looking at distant objects, they stay the same distance apart
> > and rotate to be closer to parallel.

> Bob:
> I agree with Chris here that following actual human eye characteristics
must
> be the way to go, however maybe the reasoning not to do so has to do with
3D
> photography in some way.

***
    Correct . . . the angle of convergence is the thing we need to simulate.
The 1:30 was just something I recall using just to get the effect close.
I think that's like looking at something about 7 or 8 ft away which covers
a lot of renderings of scenes that contain just one abstract object . . ..
However, when simulating a real world scene, it does become important
to maintain the perspective.

Again . . . we see that simulating what the human eye/brain is capable of
doing is quite a task, eh?  The eye can adapt to different ranges of color,
brightness and focal distances automatically . . . we aren't even aware
that it takes place. However, when using a camera or POVray to render a
3D scene, we have to pick one spot and get the settings right or it won't
look like what the eye/brain sees.

  I have done 3D photograpy for 30 years and with the camera, it's pretty
straight-forward .  I pick the object of interest and take the left and
right
shots ensuring that when I make the shift ( about 3 inches in this case . .
.
just simulating the eye separation) that the object of interest is what I
keep centered.  The result is . . . the difference in the angle from the
left
and right view will accuratly simulate the angle that the eyes use to focus
on the object of interest.  If focused at infinity, it would be zero, and as
the
distance from eye to object decreases, the angle of convergence increases.
The brain, which has learned through experience, uses both this angle to
focus the eye and the relative horizontal shift of the objects in the view
to
determine depth and perspective. And . . . notice that the object of
interest
will not shift while objects closer will shift left and those farther away
to the right.  Well . . . it's all relative . . . but focusing on the object
of
interest is important.

  With POVray . . . we have to caculate the proper camera shift.  There is
no need to calculate the angle . . . just the relationship of distance to
shift.
(Although that would be useful if we are rotating the camera around the
object by degrees - simple trig function, eh?)
  If the object is 10 foot away and we use 3 inches for eye separation, we
will want to shift our camera 3in/120in or 0.025 units for every unit of
distance to the object. For 30 foot . . . 3/360 or 0.00833 units of shift
per
unit of distance.  Of course you can deviate from this formula to enhance
the effect but . . . at some point it will no longer appear realistic.
  And  . . . using focal-blur to simulate depth-of-field would also enhance
the effect.  Keep the target object in sharp focus. That's the reason I
rotate
around the object of interest when using the camera for 3D photos . . . the
brain uses that convergence data to tell the eye the distance to focus at
and we want to simulate that too to make it as realistic as possible.

Have attached a 3D pic I found on the web . . . on a site for MathArt . . .
http://www.lvcablemodem.com/esullivan/webquest.html

This is a very nice example of what can be done.

Y
cross-eyed and unfocused . . . SOP ;)

Post a reply to this message

Attachments:
Download 'mstytwrp.jpg' (14 KB)

Preview of image 'mstytwrp.jpg'

"yooper" <Out### [at] huntelnet> wrote in message
news:3a9954cf@news.povray.org...
>
>   I have done 3D photograpy for 30 years and with the camera

That's a long time :-)  I've only been doing 3D photos for about a half
dozen years.
Some pictures of bridges and dams had a separation of around 1 yard or 1
meter with the bridge/dam center points being something like 1/4 mile away
or more.  That would be about the 1:300 ratio (well a little more) which
makes the whole scene appear to fit within a football-sized area.  Otherwise
there wouldn't be enough 3D effect to do any good.

> (Although that would be useful if we are rotating the camera around the
> object by degrees - simple trig function, eh?)
>   If the object is 10 foot away and we use 3 inches for eye separation, we
> will want to shift our camera 3in/120in or 0.025 units for every unit of
> distance to the object. For 30 foot . . . 3/360 or 0.00833 units of shift
> per unit of distance.  Of course you can deviate from this formula to
enhance
> the effect but . . . at some point it will no longer appear realistic.

I've tried both parallel and convergence and I'm not sure which is better,
but I think there's usually an artificial look to the parallel way.

Bob H.

Post a reply to this message

Bob H. <omn### [at] msncom> wrote:

> I've tried both parallel and convergence and I'm not sure which is better,
> but I think there's usually an artificial look to the parallel way.

Well, I did some tests this weekend, and I didn't really like the results. 
Of course, I don't have any experience in this particular area.

So, I'm not certain what I'm going to do.  Maybe I'll animate a walk
sequence through the doorway and looking back.  First I want to finish the
pic, though.

Geoff

Post a reply to this message

Another new version:

http://darwin.cwru.edu/~wedig/summer/summer.png

Also, I've started animating it.  There's a prototype at:

http://darwin.cwru.edu/~wedig/summer/summer-prototype.m1v

and the frames I've done so far at:

http://darwin.cwru.edu/~wedig/summer/summer.m1v

It's only the first 20 frames, though.  It takes over 11 hours for each
frame on the hardware I have, though I have access to a Mosix cluster that
I'm using, so I can do 10-30 frames at once (depending on how much of a
glutton for resources I want to be. ;) )

So, what do folks think?

Geoff

Post a reply to this message

The snow looks good and the bird is a nice addition, but the summer looks
much too dark in relation to the winter.  The light coming through the
door is quite bright, so the scenery should be brighter too.  

For some reason i could not view the movies, media player loaded them, but
no viewable content.

Christoph

-- 
Christoph Hormann <chr### [at] gmxde>
IsoWood include, radiosity tutorial, TransSkin and other 
things on: http://www.schunter.etc.tu-bs.de/~chris/

Post a reply to this message