Hi,
After some more hours of coding, I have created a autopano program
that can successfully create single row panos (hugin project/PTOptimizer files)
from normal rectilinear input images. Its quite fast as well, takes
~ 2 minutes for a 5 image (input size: 1200x1600) panorama. While it looses
many points during the tracking phase, it usually turns out to produce
nicely registred panos :)
Since the licencing issues of the really nice KLT tracker code
from Birchfield, http://vision.stanford.edu/~birch/klt/,
is a bit unclear, I'm not adding it to the CVS, but I'll provide a tarball
with the sourcecode on my homepage. The author hasn't answerd my questions
yet (well I just started coding autopano yesterday night ;). So this version
is just for research purposes.
I'll put it up at:
http://wurm.wohnheim.uni-ulm.de/~redman/hugin/hugin_autopano-0.4.tgz
it still contains the binary, so maybe you can get away without recompiling.
Just build as usual (under linux). autopano and needs the usual vigra support, and
additionally fftw (I belive version 2 is needed). It doesn't use the vigra
fftw interface, so if vigra wasn't compiled with fftw, it doesn't matter.
no time to do windows makefiles or binaries, since I should work on my
thesis instead ;)
usage:
island.wh-wurm.uni-ulm.de$ ./autopano
./autopano: create a .pto file with control points, from a given image series
It uses phase correlation to estimate an initial image shift
which is then used by a Kanade-Lucas-Tomasi Feature Tracker
to select corrosponding points.
currently assumes single row panos, where the images have been taken from
left to right
Usage: ./autopano [options] pano.pto image1 image2 image3 ...
all images must be of the same size!
[options] can be:
-n number # number of features to track, default: 100
-o number # overlap between image in percent, default: 50.
doesn't need to be precise, 10 or 20% higher than
the actual overlap might even work better
-v number # HFOV of images, in degrees, Used for images that do
not provide this information in the EXIF header
default: 40
-s number # feature tracker window width and height, default: 7.
increase if not enough points are tracked.
please try it and tell me what you think about it.
The default tracker window is a bit small, for big images, try increasing
it with the -s option
load the resulting file into hugin, or run PTOptimizer on it.
ciao
Pablo
--
http://wurm.wohnheim.uni-ulm.de/~redman/
Please use PGP