On 27 September 2011 21:41, Bruno Postle <bruno(a)postle.net> wrote:
On Tue 27-Sep-2011 at 01:14 +0200, Kevin Kofler wrote:
> Bruno Postle wrote:
>> Note that Hugin panorama stitcher now has a fully working
>> patent-free replacement for autopano-sift-C (you need to reset your
>> Hugin preferences to enable this), so there isn't really any need
>> for rpmfusion to carry this package after F16.
> What makes it patent-free?
autopano-sift-C has an implementation of the SIFT technique, and the
alternative panomatic implements SURF. Both algorithms are apparently
patented, so neither tool could go into Fedora.
This was a problem, so Pablo d'Angelo (the original author of Hugin) wrote
cpfind, which uses code from panomatic but has an entirely novel feature
Thanks for the tip. I discovered that I had to re-import the default
control point detectors to get the Fedora 15 ones (I've just upgraded
> Is the quality of the output comparable?
Yes, the cpfind in Hugin 2011.0.0 is much more reliable than
autopano-sift-c, but not necessarily faster. Unfortunately 2011.0.0 didn't
make it into F15, though it will be in F16.
Trying cpfind on a few panoramas with 2010.4.0.854952d82c8f I've found:
2 where results were acceptable (similar to sift)
1 where results were worse (a very dark image which neither could
automatically handle, but sift was able to cope once some manual
points were added).
1 where cpfind performed better (correctly connected end image of 5,
whereas sift anomalously connected it to the wrong image).
If 2011 is an improvement then on this limited evidence I couldn't complain.
Further question then while on the topic: libpano13 is apparently
compiled with a 179 degree limit to avoid a patent on hemispherical
coverage (aren't patents great?) does anyone know if that limit been
circumvented? (The actual patent, not the number in the source code. I
don't even know if it's much of a problem for single row panoramas