> >The first and last image (representing the largest spatial shift from the 
> >stars rotating in the sky) are still off by 5 or 10 pixels. 
>
> I would first check that you are optimising enough lens parameters, at 
> least barrel distortion and angle of view (in addition to positions).
>

Brilliant Bruno, that was exactly it: including barrel distortion solved 
it. (Optimizing for View as well made it a little bit worse, which I guess 
makes sense because the camera was on a tripod... ?) I still needed to use 
my custom cpfind options as described above, but the alignment and 
subsequent median_image worked great.

Excited, I turned to a second stack of star pictures, but for some reason I 
can't figure a way to make cpfind find good control points on them.

They are also underexposed, but very similar to the first stack, and there 
are still plenty of medium-white stars, and in fact there is less low-level 
noise in these images than the previous stack. I understand that these 
dark, almost featureless photos are not the typical use case for cpfind, 
but maybe there is something I can do to make this work?

In case anyone has a sec to take a look, here is a demo project with just 
two of the images and an exclusion mask over the foreground elements:

http://caseyconnor.org/pub/image/hugin_demo.zip

I think you'll find that the default CPfind settings will find no points. 
With my custom cpfind options it generates 4 points, but two of them are 
obviously wrong. I've been tweaking the cpfind options and with this:

--fullscale --sieve1width 10 --sieve1height 10 --sieve1size 100 
--sieve2width 5 --sieve2height 5 --sieve2size 5 --kdtreesteps 100 
--ransaciter 100 --ransacmode rpy -o %o %s

...I can generate 8 points, 2 of which are bad. That's pretty good, but 
when I try it with all the images (there are 63 in this stack) it just 
doesn't find enough points. Very many pairs of images have no points at all 
and the final alignment is poor.

Anything I can do to get better feature matches?

I tried to pre-process the images to be brighter, find the CP's, with the 
idea of then swapping the original images back in before stitching. When I 
did that I couldn't find any CP's at all, which surprised me.

Any thoughts on what to do? Thanks!
-c

(P.S. it'd be great if there was a --starfield mode that just looked for 
bright points to align!)

-- 
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/f45a41eb-0d59-45c6-97c1-d2cb1e87650c%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to