Re: [photo-3d] The Stereoscopic Society Annual Competitrion

["John A. Rupkalvis" <stereoscope@xxxxxxxxxxxxx>]

----- Original Message -----
From: "William Gartin" <william_gartin@mac.com>
To: <photo-3d@yahoogroups.com>
Sent: Wednesday, February 14, 2001 9:07 PM
Subject: Re: [photo-3d] The Stereoscopic Society Annual Competitrion


> on 2/14/01 12:07 PM, Paul Talbot wrote:
>
> > "David W. Kesner" wrote:
> >>
> >> Paul Talbot writes:
> >>
> >>> My WAG: he used a
> >>> very small stereo base.  But I have no macro experience and
> >>> you do, so I presume you can't make the math work with the
> >>> MAOFD formulas, right?
> >>
> >> I have no idea what the stereobase was on this image, but it
> >> was large enough to achieve good stereo.
> >
> > My guess: only a very small stereo base is needed when there is
> > something that close to the camera and the scene extends a long
> > way.
> >
> > Paul
> >
> >
> >
> What would be considered an appropriate separation angle in degrees? If
you
> know that, then all you should need is the distance to the subject to find
> the appropriate separation, right?
>  --
> William Gartin <william_gartin@mac.com>


No.  The optimum separation angle (and the resulting parallax) are variables
that cannot be quantified for all subjects.  Each subject is different, and
several factors enter into the judgment as to what is best for a specific
stereo image.  Things such as contrast, color, association, sharpness,
relationship of other objects in the image, etc., etc., all must be
considered in the evaluation of the most appropriate settings.  Even whether
convergence is parallel (shift lenses or shift mounts) or toed in.

You must consider that the eye/brain system does not instantly take in an
entire image simultaneously, but rather "scans" the image.  If an object has
nothing between it and the background, much narrower parallaxes are required
than if there are objects in between.  With intermediate depth references,
the eyes can track from one object to another, rather than "jumping"
directly between foreground and background, thus permitting wider parallax
differentiation between the nearest and farthest points in the image.

I use an extreme example to illustrate this.  It is possible to "create" an
image such that nothing is changed except the subject, and in one instance
it will work, and in the other not.

An actor is standing in the foreground, say about two meters from the camera
(Stereo Realist, dual cameras, whatever).  The scene is in the Mojave
Desert, with nothing else in the picture except the mountains on the
horizon, several miles away.

Regardless of where the cameras are converged, it will not be possible to
view the image without severe eyestrain, and a "splitting" of either the
image of the mountains, or the image of the actor (depending upon which you
concentrate on).

Then, do not move the camera or make any changes to your photographic
procedure.  Simply add another object (a prop such as an artificial cactus
will do nicely, or just drive an automobile into the scene).  This
"reference" object should be about four meters from the camera.

Now, the image will be much easier to view.   Same close point and far point
parallaxes, everything else the same.   Stereoscopic imaging is extremely
subjective.

Many people have spent a lot of time creating mathematical tables.  For the
most part, these are most useful for the designers of cameras and other
imaging equipment.  The stereo photographer would be far better off spending
that time in making pictures and learning by experience what factors affect
the optimum images under what combination of circumstances.  Visual
viewfinding systems (reflex cameras, video camera or camcorder viewscreens,
etc.), as well as studying the resulting images, can be very educational.

JR