I'm creating panoramas using lenses of different focal lengths.  My
technique is to take a strip of photos with a shorter focal length
lens (say, 28 mm) and within that to take another strip of shots with
a longer lens (e. g. 50 mm).  This is useful for landscapes, for which
having a lot of detail in the midground and horizon areas is
important, but not so much is needed in the sky or in the extreme
foreground (which typically frames the shot or would be out of focus
anyway).

Hugin (7.0) in general doesn't have any problem with this, but there
are three specific issues:

1) Fitting control points is problematic.  Autopano actually does a
   decent job of finding points, but they're not all that accurately
   positioned, and manually creating control points and fine tuning
   them is much more difficult.  It looks like what happens is if a
   control point is placed inside or outside of a corner or an edge,
   the placing algorithm doesn't take into account the difference in
   focal length.  For example, if I place a point 4 pixels from a
   corner (vertex), the autoplacement algorithm places it 4 pixels
   from the corner in the other image, even if the focal lengths are
   different.  Fine tuning is useless; I have to manually place the
   other side of the control point.

2) Enblend doesn't know to give preference to the shots taken with the
   longer focal length lens (i. e. the ones with more detail).  The
   actual result depends upon the order of the images: if the short FL
   ones come first, they all get used, and then enblend ignores the
   long FL ones since they aren't actually needed, while if the long
   FL ones come first, the end result is a mosaic of long and short FL
   pieces.

3) It's possible to move single images up or down by means of the move
   up and move down buttons in the Images tab, but not multiple images
   at a time (if more than one image is selected, the move up and move
   down buttons are grayed out).

The way I work around the second issue is as follows:

* Use Hugin to generate discrete TIFF files.

* Run enblend manually twice, once on the long FL images and once
  on the short FL ones.

* Run gimp on the short FL composite.

* Load the long FL composite as a separate layer above the short FL
  composite.

* Crop the combined image to remove the areas with no image (I wish
  hugin could do that!).

* Create a layer mask for the long FL layer using the alpha
  channel.

* Show and edit the layer mask.

* Use select by color on the layer mask (which is all black or
  white at this point).

* Shrink the selection (of the layer mask!) by 50 pixels

* Feather the selection by 50 pixels.

* Invert the selection.

* Bucket fill the selection with black (transparent, since we're
  editing the layer mask).  This yields a blend zone 100 pixels
  wide.

* Un-edit and un-show the layer mask.

This yields a remarkably smooth blend zone with typical images, but
it's a bit more work than I'd like to do manually.  In particularly
bad spots, such as crossing sparse foliage where the images aren't
well-aligned due to wind or parallax error (I don't have a Nodal
Ninja or the like), the transition is visible -- if you know
exactly where to look -- but not objectionable.

The first time I did this, I cut a hole out of the short FL composite
(so I could pick exactly where I wanted the transition) and used
enblend to merge them.  That actually seems to yield inferior results
-- most of the time -- to the rather simple-minded approach above.
That seems to be because enblend uses too narrow a transition region
in areas with detail, but that isn't always the right thing to do --
if the detail is highly repetitive with low information content (such
as pine needles), a broad transition zone is more appropriate.

This does yield some ghosts where things aren't aligned, but in the
ones I've shot so far, those ghosts are typically individual leaves or
twigs that aren't really much of a problem even if you do know exactly
what you're looking for.

Automating this workflow wouldn't be that hard, I'd think.

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"hugin and other free panoramic software" group.
A list of frequently asked questions is available at: 
http://wiki.panotools.org/Hugin_FAQ
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at http://groups.google.com/group/hugin-ptx
-~----------~----~----~----~------~----~------~--~---

Reply via email to