PanoPoints

[JD Smith]

On Sun, 2003-10-05 at 12:57, Pablo d'Angelo wrote:
> Hello JD,
> 
> I just tried to run panopoints, it starts, but as soon as I click on
> the "file open" icon, a warning message box appears:
> 
> "Warning: Can't open script: Gtk::Button=HASH(0x8385120)"
> 
> I'm using debian, with the following packages installed:
> 
> libgtk-perl-0.7008
> libgtk-1.2.10
> perl-5.8.0
> 
> Any idea what might be wrong?

Thanks for giving it a try, Pablo.  Since I always call panopoints like:

panopoints myscript.txt

I hadn't caught even this simple bug.  Goes to show you that a user base
is critical.  Anyway, it's fixed now.  The easiest way to get it is with
CVS; see http://sourceforge.net/cvs/?group_id=86324.  Nota Bene: there
is a delay between CVS commits and the Browse CVS data (which you're
probably aware of, but just in case...).  Note that panopoints works
best on pre-existing script; you might try it on one of your old
projects.

There are some simple directions and a key-shortcut reference at:

http://panopoints.sourceforge.net/

> I don't think I'll port hugin to another toolkit, but I interested in
> seeing trying your program, to smoothen the hugin workflow.
> Contrary to PanoPoints I want to add some image processing /
> semi or fully automatic control point creation, not just a a smart UI to
> make picking points easier.

Have a look.  I think I already made a significant improvement to the
workflow by just averaging existing control point offsets, and using
that average offset to automatically place the other point of the pair. 
For example, once you have a marker pair or two, position the cursor on,
e.g., the left image, and hit "p".  A marker appears in the right image,
typically very close to its final destination.  You can use the mouse or
arrow keys (possibly with Shift accelerator) to move the other member of
the pair into place.  This obviously works well for rectilinear images,
but less well for fisheye or others, where a simple linear offset isn't
quite right.  The back-up, if this predictive calculation goes awry and
puts the other point marker off the image, is to do it the "old
fashioned" way ala PTPicker: use offsets relative to the image portion
visible.

> It would be nice if you considered contributing to hugin, but I see that
> wxWindows/c++ is quite different to perl/gtk.

I have to admit that I didn't even get around to trying hugin because of
the wxWindows toolset requirement.  I worry that others will be lazy
like me: tired of the proliferation of widget toolsets.  I guess
wxWindows gets you cross-platform compatibility, but there are already
so many Windows PanoTools front ends, that it seems the Unix world is
the place to make a stand.  I do promise to give it a try at some point
soon; is there a quick how-to page?

> > If I ever dug deeply into Panotools, I'd probably try
> > to rewrite the core stitching library to make it faster, more plug-in
> > friendly, and fully open (e.g. in the way PTOptimizer/PTStitcher are
> > not).
> 
> Yep, they are not very nice, I'm using some functionality from them in
> hugin, but one can see from the code that H.Dersch is a physicist ;)

Do you have the source for PTOptimizer/PTStitcher (and not just the
libpano library)?  I haven't seen it in the wild, despite it's use of
libpano, and GPL-mandated source distribution.  Helmut hasn't answered
my pleas to give the source.  The old linux binaries for these two still
in use will not survive major changes to the executable format, etc.

> Do you have any plans to start something like that? I'm very interested
> in it as well (actually I'm already putting bits and pieces of the
> formulas into hugin..). Some time I'll have to complement my panorama classes
> with remapping functionality, to do proper pattern matching between
> different images, for a faster preview etc.

Not planning on it, but have always thought it would be fun.  I also
have some new ideas on how to cut down the noise associated with
interpolative re-sampling based on some techniques for astronomical
imaging.  I'm an astrophysicist, by the way, but I'll let your
disparaging physicists comments slide ;).

Thanks again,

JD