Isn't the answer to the situation shown in your diagram, where there is
an isolated area of crumpling (and yes, I know it's only a diagram, but
the situation does arise in reality), to take an extra photo from
directly above the crumpled area and looking straight down on it and
then force the use of that version with an Include mask. It's probably
better to avoid control points in the actual crumpled area because of
the parallax they embody.
The situation is thus analogous to taking a photomosaic of a church
elevation, where, for instance, it's vital to take images of protruding
features like buttresses and receding features like windows from
straight-on and then select those versions with masking. I think there
was an early tutorial on the topic.
Roger Broadie
-----Original Message-----
From: bugbear <[email protected]>
To: [email protected]
Sent: Thu 28 May 2015 09:07:27
Subject: Re: [hugin-ptx] Mosiac model - convergence, accuracy?
Roger Broadie wrote:
Hello Paul,
I thought your panorama was pretty good, but I don’t think you have
fully overcome the problems of parallax resulting from creases and
folds. The main cause, I suspect, is the camera tilt of 6 deg plus or
minus from the vertical that the pitch values of the optimiser imply.
I think that 3D items can cause trouble, even if the camera is shooting
perfectly vertical. While the camera may be perfectly
vertical w.r.t item directly beneath it, it is (of course) at an angle
to items at the edge of its FOV. When the camera
is translated in X,Y, (possibly Z), the item is now at a different point
in the new FOV.
Diagram attached.
So while I would not normally expect to include the Plane Yaw and
Pitch parameters in a truly planar photomosaic, it did seem worth
trying them here. They do seem to help.
Yes - if the map is NOT a true plane, but compose of multiple planes (or
at least may be thought of as such)
the sub-plane parameters may help, at least "on average".
I attach panoa.pto, based on your file, which takes that approach and
also includes some tuning of the control points and fuller lens
optimisation. But I hung on to the h and v points, because I think
they are one of Hugin’s strengths. I have resolved some remaining
glitches, either with added control points – automatic control point
finders seem to find otherwise featureless areas of grid-lines
difficult to deal with – or with masking, but have not tried fully to
resolve problems with the
grid-square P25, where the different shapes of the square as captured
in e.g. images 17 and 23 illustrate graphically the problems of the
parallax present in places in your images. The end result yields an
average error of 1.2 pixels and a maximum of 3.2, for a panorama
matching that of pano.pto in size.
Thanks for the observation - I'll go look!
Broadly, I agree with Terry Duell (Wed, 27 May 2015 at 10:13:42 +1000)
that the mosaic mode is working OK. He has given us a pto file (Wed,
27 May 2015 at 12:34:56 +1000) that does not use the Plane Yaw and
Pitch parameters. It has an error of 0.5 average and 2.9 maximum, but
is for a panorama of only about half the width of yours. Still pretty
good, though, even at the full size.
Agreed - thank for you effort and thoughts.
My problem of how to best capture a 1816 map remains, of course. :-)
BugBear
--
A list of frequently asked questions is available at:
http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/hugin-ptx/5566E86C.5090402%40ogea.freeserve.co.uk.
For more options, visit https://groups.google.com/d/optout.
