On December 27, 2010 08:32:49 am Rogier Wolff wrote:
> Control point finders are apparently quite likely to find a bunch of
> control points in the foreground.

no - there is no particular likelihood.  they find CPs indiscriminately on the 
whole surface of the picture without any depth information.

CP finding could be improved if we could add depth information to the process 
and either mask out the near from the far; or, better, qualify its CPs with 
its depth.


> Which especially with hand-shot panos are quite worthless.

true.

 
> So after one run of finding control points and aligning the pano,
> would it be possible to run a CP finder on just the overlapping image
> parts?

I see no benefit in this.  There is no guarantee that in the part of the image 
that is overlapping depth is distributed any better than in the rest of the 
image; and parallax would have to be really massive to keep the identified 
foreground features out of the overlap area.

Yuv

Attachment: signature.asc
Description: This is a digitally signed message part.

Reply via email to