P3D Re: Instant Anaglyphs

[Jim Crowell <crowell@xxxxxxxxxxxxxxx>]

At 6:12 PM -0700 8/11/98, Tom Deering wrote:
>>
>>Well perhaps I was unclear about one of the camera details.  I
>>intended to have both the left and the right pinhole expose the same
>>piece of film with no septum or other separating device.  The image
>>formed by the red and green pinholes should merge to form a single
>>image.
>
>Nope.  Here's some illustrations that show the kind of pictures this device
>would make:  http://www.deering.org/doomed.html
>

Not _completely_ doomed, just mostly.  Here are the three reasons from
Tom's web site:

>The idea won't work for three reasons.
>
>1.     This arangement will yield color separations, not an anaglyph. Note
>how the red stripes are >removed from the red portion of pict 2. This is
>what a color separation is all about.

As a couple of people have pointed out, this is not entirely accurate.  The
more accurate statement is that it won't work for scenes with arbitrary
colors.  The resulting pictures are called full-color anaglyphs, and the
fact that they don't always work real well hasn't stopped people from
publishing books full of them (e.g. some of the books in the Nature Company
"____  in 3D" series were full-color).  Sometimes it works, more often it
don't.

> 2.    The two halves of the image will be separated by the interaxial
>distance, the measurement between the two lenses. As objects get closer to
>the lenses, they will be separated further. This is very different than an
>anaglyph, where the two images are on top of each other.
> 3.    The two halves will be in reverse order, due to the way the lenses
>invert images. Unless you use prisms, you can't view these images in
>stereo.

These are basically the same reason, and this is the kicker.  When you
mount a stereo slide, you actually do a left-right reversal of each film
chip (i.e. you turn it over) while keeping the left one on the left & right
one on the right.  I think the Q-DOS lens accomplishes this in taking the
picture.  There's no way to do this when the two views are already stuck
together in a single image.  So as Tom says, you'll end up with near
objects widely separated in the image.  If you were to try to view the
image taken through the left hole with the left eye, you'd have to diverge
beyond parallel to see anything that was nearer than infinity in the
original scene.

You might be able to get away with reversing the filters when you view the
image.  That would bring infinity in fairly close (your right eye would be
looking at something directly in front of the left, and vice versa, so
infinity would be in front of the image plane), but at least everything
should have the proper depth order...

-Jim C.

-------------------------------------
Jim Crowell
Caltech Division of Biology
Mail Code 216-76
1200 E. California Blvd.
jim@xxxxxxxxxxxxxxx
http://vis.caltech.edu/~jim/Home.html


------------------------------