[ptx] automatch - some progress picking control points

[Bruno Postle]

On Wed 14-Jan-2004 at 07:55:41AM -0500, Ed Halley wrote:
> On Wed, 2004-01-14 at 06:11, Bruno Postle wrote:
> 
> > My latest version tries to validate each set of control-points by
> > trying to build them into a network - It has a bug where sometimes
> > all the points it finds are correct and sometimes all the points are
> > wrong.
> 
> Are you trying a Delaunay triangularization?

Never heard of it until now :-)  Though I don't see at first how it
can be used to recognise patterns.

I'm building a network by taking the two highest-scoring point-pairs
and seeing if any of the other lower-scoring point pairs can be used
to make a triangle that is the same dimensions in both images.

(This method fails when I've picked the wrong initial points)

-- 
Bruno