[photo-3d] Re: Algorithm Wanted

["Bruce Springsteen" <bsspringsteen@xxxxxxxxx>]

I love it!
Dok T comes through with a solid shot over the wall:

> Question:  Can you identify homologous points?
> If you can identify the homologous points then the 
> problem is simple. :)
> 
> You position one pair randomly "over the other", select 
> a number of homologous points and compute the displacement 
> vectors.  You sum up all the vectors together.

Well, what could be simpler!? :-)  Is this possible for
non-math majors?  You mean randomly pick one pair of
homologous points and lay one over the other, right?

> You then have the following problems:  
> 
> 1. Shift one pair with respect to the other to minimize
> the total displacement vector (two variables).
> 
> 2. Rotate one pair with respect to the other to minimize
> the total displacement vector (one variable).

Everybody got that?  Need some clarification?  A demo maybe,
George?  I'm a bit unclear, but think I have the general idea.
No push-pins involved at this point?

> The logic behind this procedure is that when the
> members of the stereo pair are perfectly aligned then 
> the total displacement vector has a minimum magnitude 
> and points in the horizontal direction.
> 
> You can then set the stereo window by finding the point
> with the smaller displacement vector.

Now that is a beautiful touch!
 
> If I understand this correctly, there are only three
> variables involved so three points will suffice to solve 
> the problem.

Right, which answers the question I just posed in response to Ken's
idea.  But will *any* three points do?
 
> What do you say Bruce?

Fantastic, if I understand it as well as I think.  Two questions:  Do
you have a handy way to do this in the real-world, that
is, how do we mechanize your algorithm and sell it?  And do my
"prints" have to be "transparencies" to accomplish this "overlay"
stuff? :-)

My high school education is straining with the effort.

Bruce