[ptx] how to do 'correct' 360 deg. panoramas?

[Pablo d'Angelo]

Hallo Mike!

Mike Runge schrieb am Dienstag, den 03. Februar 2004:

> hope that works like expected. You'e right, constraints for v/p would be the
> preferred solution for that issue.
> 
> For panos that are not closed I never optimize the v parameter. I found out,
> that there are a lot of acceptable solutions for the same pano with various
> values for hfov. And I found no rule to steer the optimizer to the 'right
> one' while optimizing v together with more than one other parameter.

Hmm, haven't experimented with that yet, but to me it seems sensible to
modify the image positions while estimating v. My hope is that more than
3-4 points per overlap lead to a better optimisation result, especially in
these cases.

> It further looks to me that equirectangular optimisation is more sensible for
> that issue than cylindrical?! But I can be wrong on that. I didn't followed
> that observation - maybe I did something else wrong?!

doesn't matter at all (except for horizontal or vertical points),
optimisation is always done with a spherical/equirectangular control point
distance.

ciao
  Pablo
-- 
http://wurm.wohnheim.uni-ulm.de/~redman/
Please use PGP