Re: [photo-3d] All it takes for 2d to 3d

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

In this example, all of the stereoscopic information already exists, since
it is a procedure for generating altitude-dividing isolines for a
topographic map from already existing stereo pairs of aerial photographs.

In the case of generating 3D from 2D, the procedure is very different.

One common approach is to first determine what actual stereoscopic
information exists in terms of time-parallax (frame-to-frame lateral shift,
displacing foreground information relative to background).

Then, other monoscopic depth cues are used to generate a depth map, which
may or may not necessarily be a reasonable match for a true stereoscopic
image.  Variables include the amount of detail used (or even available) the
clarity of the cue (such as focus, edge effects, density, etc.), and, as
already alluded to, any image parts that are obscured.

Obscuring of parts of the image (called occlusions) are not all bad.
Sometimes, as when part of an image is obscured in one frame but revealed in
another, this can actually be used as an additional depth cue.

As has already been pointed out, it certainly is not simple.  As I have
pointed out before, what you can do and how much you can do, are directly
related to how much time you have and how deep your pockets are.

JR

----- Original Message -----
From: "Joshua N. Rubin" <jnr@xxxxxxxxx>
To: <photo-3d@xxxxxxxxxxx>
Sent: Tuesday, October 31, 2000 7:39 PM
Subject: Re: [photo-3d] All it takes for 2d to 3d


> > There are many examples I can think of as well. I did not say it would
be
> > simple and it most certainly will not be. However, the possibility still
> > is a very real one. Is it an automatic process, NO. It would be labor
> > intensive as technology develops.
> >
> > However, if the past is any guide, if there is money to be made, the
> > process will also be created. But most of what we have today also would
> > be debated both ways as well. All it takes is Power, Time and Money and
I
> > see no problem creating a system that can recreate a missing view. No,
it
> > will not be something any of us will have on their desktop.
>
> WARNING - BALD CONJECTURE AHEAD
>
> I think that probably the most advanced means for doing this has been
> developed for extracting depth from aerial stereopairs of terrain.  Is
this
> called aerial stereophotogrammetry or something like that?  Whatever it's
> called, I imagine that it is or could be largely automated because many of
> the geometric variables would be known or fairly tightly controlled.  I
> would imagine that a human operator might have to match up a few
homologous
> points between the scanned images but that a sophisticated pattern
> recognition program could do the rest.  Of course, this is a special
> situation because only a very tiny part of the terrain visible in one
> perspective would be obstructed in the other and there are no surfaces of
> interest that are not visible in either view.
>
> DISCLAIMER - I AM NOT AN AERIAL STEREOPHOTOGRAMMETRIST, NOR DO I PLAY ONE
ON
> TV.
>
> Josh
>
>
>
>
>