T3D Re: Stereo Math

[john bercovitz <bercov@xxxxxxxxxxx>]

> I am interested in quantifying the amount of total parallax in a
> scene, foreground to background, as a way of talking in specific 
> terms about some pretty subjective impressions.

I like this idea.  Anyone can measure the total parallax in an
existing stereo pair even if the pair is already mounted just by
subtracting the separation of the nearest points from the separation
of the farthest points.  

I'd add two refinements (you might also call them complications 8-).

First, I like to see it in angular terms so I would divide the parallax
by the focal length of the lens to get a faux-angle.  [Macros do get a 
bit more complicated because the lens position is moved substantially 
relative to the film plane.]

Second, I think it would be good to note the highest individual case 
of parallax.  Let me explain what I mean by that.  If you take a 
picture of a flower which is 7' away against a mountain which is 
essentially infinitely far away, then the parallax of the flower
against the mountain will be 1.2/35 = .034.  If instead, you take a 
picture of a well-grazed featureless prairie extending from 7' out 
to infinity, the total parallax is still .034 but there is no
individual case of parallax amounting to much more than 0 because 
nothing occludes anything else.

John B


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