Hi,
I'm looking on some advice on stitching together a set of images where a
few images might have no key points, but where I do know their position
(without reason).
I previously described my setup in a mail to this list titled "Field of
view, crop and canvas size" (see [1] for a part of that mail). I have
since adopted pto_template and cpfind --prealigned in my pipeline.
In my project, we shoot a total of 16 photos in a 4x4 grid, numbered as
follows:
01 02 03 04
08 07 06 05
09 10 11 12
16 15 14 13
The stitching process is headless / automated and there is no room for
human interaction. To prevent bad stitches, I have written code that
calls and parses 'checkpto foo.pto --print-output-info' to ensure the
min/mean/max error are within acceptable thresholds and all the images
are connected. The code also checks the exit status of checkpto (it has
to be zero).
Occasionally some photos (say one or two photos in the 13 - 16 range)
contain almost no useful content / features for control points. In this
case, the checks will usually trip up because unconnected images are
found. Assuming that I can write some code that the detects if this is
the case for a set of images, I was wondering what my options would be
with Hugin.
From my perspective, ideally I could still produce an image with these
16 shots included/stitched and accept that there will be visible seams -
given that there is not useful content this likely would not matter that
much.
Not including the images without content in the stitching process is an
alternative option, but I am hoping to not have to do that - in the odd
chance that some of the images do contain some useful content.
Which brings me to my two questions:
1. Is there a way to naively define an offset for an image with respect
to another image when there are no control points? I suppose I could
write an estimated TrX and TrY in the 'image lines' portion of the pto -
but without any control points, I suspect the images will still be seen
as 'not connected'. Should I insert some synthetic (but hopefully semi
accurate) control points? Will that work, or will those get discarded by
say cpclean?
2. Is there a way to use the existing command line tooling to 'remove'
or 'disable' an image in a project file? (Disabled in my book would mean
that checkpto would ignore it and it also wouldn't be included in the
final result)
With (2) I could temporarily disable some images and run checkpto and
confirm that the other images are all properly linked and (1) would
allow me to then place the remaining images with a known offset (again,
accepting that they'd be slightly misplaced/offset).
If I do end up programatically inserting some 'mostly accurate' points
in the .pto files, is there anything that I need to keep in mind when it
comes to optimising the .pto file? Like, should I tell autooptimiser to
not optimise for the synthetically inserted images?
Please see this link for an example of what the 4x4 grid looks like [2]
(this is not a stitched image, just some scaled down images pasted in a
grid with a black boundary separating them).
Finally, thank you very much for hugin and panotools.
Thanks in advance for any insight/help,
Regards,
Merlijn
[1] A few quick details: the images are made at a high magnification and
high resolution (151PM). The stitching is part of an effort to digitise
microfiche cards and the stitching is a fully automated process - so
people aren't supposed to say launch Hugin and fix some problem(s). The
output images are about ~1.3 gigapixel grayscale images.
[2]
https://archive.org/download/micro_IA40386401_0002/micro_IA40386401_0002_itemimage.png
--
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/1c2bac21-9e21-460e-a511-13b24c3e3637%40archive.org.
