SL3D

[P3D <PK6811S@xxxxxxxxxxxxxx>]

Chuck Hassen wrote:

>The thread on SL3D imaging is very interesting to me; I just had a comment
>regarding the principle of operation.  When Paul points out that "The f/1.8
>aperture is great because the effect is more pronounced as depth of field
>diminishes" I wonder if maybe the cause of the more pronounced effect is a
>greater average separation between the red and blue halves, not so much the
>diminished depth of field.  The separation of left and right channels of the
[...]

The two effects (dof and 3d) are both directly related to the size of
the aperture.  Following is one of my earlier posts on the subject:
(aplogies for the repeat, but there seems to be new interest)

-------------------------------------------------------------------------
Let's start with a pinhole.  A pinhole is said to have infinite depth of
field, because light rays coming from any and all points in the scene go
straight through one hole to the emulsion.  No matter how far away the
object, and no matter how far away the emulsion, a viewable image will be
formed.

Now add a second pinhole to the right of the first.  It will also form an
image but this one will be shifted to the right of the first image.  If
there is no overlap between the images, a stereo pair is made.

If we put a prism behind the second pinhole we can shift its image to 
overlay the image made by the first pinhole.  The great benefit of this
is to increase brightness in the combined image.  The downside is that
since the two pinholes do not 'see' exactly the same scene, being offset
horizontally, there will be some slight blurring.  Unlike the color 
dispersion problem, this one cannot be overcome by a using a fancier prism.
The only help is to keep the two pinholes close together.

Also, a new problem is introduced.  The two images will overlap exactly
only for limited combinations of object-to-hole and hole-to-emulsion
distances.  This can be overcome somewhat by adding a 'focusing' mechanism
which brings the two together for an object at most any distance we want.
We just can't do it for all objects at all distances at the same time anymore.

Reality Check: the difference between the two images, and the amount of 
shifting that occurs may be so slight that we can really capture 
recognizable images of objects within a _range_ of distances with
success (depth of field).

Now add a zillion pinholes (cut out a big circle around the first pinhole).
Install a zillion prisms (a lens will do).  Box it in plastic and sell
a million of them :-)  Unfortunately depth of field may suffer noticebly in
this setup.  So install a variable iris to get back to the single pinhole when
needed (you can charge extra for this model :-)

By using multiple components and combining glass of different refractive
properties many problems can be overcome.  But one still remains.  Each
'pinhole' really does see the scene from a different perspective.  I'd
call this 'perspective distortion', but that term is already in use for a
another problem.  Anyway, this is a blessing in disguise for 3D'ers, since
we can theoretically and practically extract separate views from our single 
lens.

It can be most easily done in red-blue anaglyph, just install left-right
red-blue filters to cover the diaphram.  But it could also be done using
polarizing filters for full-color capture, unfortunately this presently
incurs a _much_ greater cost in the rest of the camera, or a specially
made film that can separately store two cross-polarized images at the
same time.

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

I also did some experiments using various masks over the red-blue filters
at the diaphram.  Smaller openings give sharper pictures.  Moving
the openings away from the middle increases red-right blue-left shading -
aking to vignetting - due to off-center aperturing, which is one drawback
to red-blue filters.  Imagine yourself standing in a large concrete
tube looking out.  As you lean to the left you see more to the right...

Paul Kline
pk6811s@xxxxxxxxxxxxxx


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