Hello Michael,
I don't understand define
horizontal and vertical lines to bias the optimisation. In fact this needs
to be fully automatic. No GUI / manual interaction at all.just the grid of
pictures with rather well known only minimally changing relative locations
and angles.
There are two special types of control point pairs where the two points
are optimised to have the same x coordinate with an unconstrained y
coordinate (or vice versa). So the pair of points defines a horizontal
(vertical) line. I can't comment on how/if to place them through the
scripting api. Defining coordinates would only work approximately if
each image has some uncertainty.
Now the algorithm should optimize the grid by fitting the overlapping
areas.
Additional question: maybe the grid is not really a rectangular grid but
the colums are slightly shifted up respectively by a known bias. Can Hugin
handle this or should the pictures be cut by a preprocessor ?
The grid should only define initial positions (I think) and guide the
control point finding such that only neighbours are considered. I would
assume slight displacements are fine for that. The actual optimisation
is unaffected.
best regards, lukas wirz
On Tuesday, February 24, 2026 at 11:16:58 PM UTC+1 lukas wirz wrote:
Hello Michael,
They are shot completely planar and all in exactly the same distance
from a
plane holding the motive (but maybe some wide-angle distortion)
So the motive is completely planar and the camera moves relative to the
object? That's not a problem, but be sure to optimise the translation
parameters.
They will overlap by - say - 200 pixels in each direction. They will show
rather sharp contours (e.g. pencil lines)
It is essential that the result is a rectangle that is perfectly aligned
with the average angle of the components.
Up to rotation that should be the natural outcome. You can define
horizontal and vertical lines to bias the optimisation towards that.
Depending on the motive, and demands, maybe fiducial markers can be
considered.
I found that it would be good to first stitch the many rows of four each,
and then stitch the column of the about 1000 resulting rectangles.
I don't see an immediate reason to split the project, unless you are
running into hardware (memory) limitations. Otherwise I'd assume you're
better off with a single project containing all photos.
Is this a viable ways to go ?
Should the (possible) wide range distortion be eliminated by some other
software before feeding the pictures to Hudin, or can Hudin easily do
this
on the fly ?
Assuming you are always using the same lens, I would recommend to
calibrate the lens / find the lens parameters and input them to hugin
(including the field of view). You can also optimise the lens
parameters together with the image positions but it's just more degrees
of freedom, ie, takes longer to optimise and it's easier to get bad
results.
best regards, lukas wirz
--
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 visit
https://groups.google.com/d/msgid/hugin-ptx/b4955129-e2fe-4eb3-8468-c57aa6009952%40posteo.net.
