[ptx] Wrong HFOV calculations for vertical panos

[Marko Mäkelä]

Hi all,

First of all, I'm impressed by the development of these tools.  I tried
PTStitcher about two years ago and have been lurking on the list since.
In the past two weeks, I have shot a couple of panoramas with my
2.1-megapixel pocket camera.

For the first panorama, I originally created the control points by hand.
For subsequent panoramas (8 pieces) I decided to give autopano-sift a try.
Even for the first panorama, it produced better results than the control
points selected by me.

I must say that the autopano-sift algorithm works extremely well for
single-row panoramas.  For multi-rows, it does not always seem to align
the images correctly.  A tree would not be shown as a vertical line,
but the bottom part could be shifted horizontally.  I didn't examine the
control points yet.  I also noticed that autopano-sift can't detect any
control points from very dark images.

I think I may have found a bug in Hugin.  If I shoot a vertical panorama
(rotating the images by 90 degrees and setting the Exif Orientation tag
to 1 before running any panorama tools), autopano-sift and the optimizer
in Hugin seem to work properly: the "Images" tab shows meaningful
angles.  However, when I calculate the field of view in the stitching
tab, it'll show way too wide HFOV angles.  The correct value would be
around 40 degrees, but it'd show at least 100.  (I can't remember, and
I can't run Hugin easily, because in current Debian GNU/Linux unstable,
libwxgtk2.4 conflicts with base X11 packages.  I had to run Hugin
without a window manager when I stitched the images.)

When it comes to usability, I think that it would be very nice to
integrate Hugin, autopano-sift and Enblend into a single process
via some library interfaces.  Currently they are loose programs,
and Hugin and autopano-sift have large dependencies.  autopano-sift
even needs X11 running, although it doesn't have a GUI.  (This may
be a bug of Mono as well.)

With best regards,

	Marko Mäkelä