OK
I moved around the images in the sequence and now the mask that was
being ignored is being applied and the mask that was being applied -
you guessed it - is being ignored. So I still have photographer's-
feet appearing in the final stitched image but these feet come from a
different shot.
In case it is any help for hunting the bug - my exact circumstances
are set out below ... It is a bit long winded but I have done a few
experimental variations in the hope of providing useful data amongst
the detail.
A. Photography
First of all, using a Canon 5D and 16mm Fisheye lens I took RAW, full
size images as follows:
- 6 horizontal shots at 60° intervals taken with camera on a
levelled tripod with VR mount
- a sky shot (again taken with camera on a tripod with VR mount)
- two ground shots taken handheld one with the photographer standing
on one side and one standing on the other
Each of these "shots" was actually three bracketed exposures:
one middle-exposure, one under exposure (-1.5 stops) and one
over exposure (+1.5 stops)
For one of the horizontal shots the camera was facing straight at the
sun I took an extra set of three bracketed shots with my thumb
blocking out the sun
(my Fisheye lens is prone to flare so this gives me a set with less
flare - in the past I have used this for patching the sky in Photoshop
after the stitch.)
I also took some extra shots at the beginning and end of the six shot
shot rotation.
B. Organisation and selection
I imported all the photographs into Apple Aperture, selected the
specific shots that I was going to use and exported them as 16-bit
TIFF files
There were 27 in all. For the three bracketed shots at the point
where the sun was mid-shot I used the one with the sun-obscured-by-
thumb for the "over exposed" shot and the ones with sun visible for
the "normal" and under-exposed. Another one of the horizontal shots
had a fragment of an unwanted moving object at one edge.
C. Masking
I took into Photoshop the TIFF files of the images to be masked.
I added an Alpha mask to:
- the three images in the bracketed set with the unwanted edge
object
- the image with the thumb blocking the sun
(I tested a theory that the bug might have been triggered by
having a mask on only one image in a
bracketed-exposure stack by adding masks to the other two images
in that stack and re-stitching
- but the mask that was ignored did not change as a result of
varying this.)
- the two bracketed sets of down shots - masking out the
photographer's feet in each case
D. Hugin
I imported 27 images into Hugin and placed them as:
- 6x3 (exposure-bracketed) horizontal shots at Yaw intervals of
60°
- 3 bracketed shots of the sky
- 2x3 bracketed ground shots
I generated control points using Auto-PanoSift-C 2.5 linking the 9
"normal" exposures and again on each set of 3 bracketed exposures.
E. Results:
Images 18, 19, 20 are bracketed exposure stack for one down shot
Images 21, 22, 23 are bracketed exposure stack for the other down shot
Images 24, 25, 26 are bracketed exposure stack for the sky shot
The masks on the images at (18, 19, 20) were ignored and the feet
from that image were included.
If I swap around (18, 19, 20) and (21, 22, 23) the masks on the
different images at (18, 19, 20) are still ignored and a different set
of feet appear.
if I move the sky stack (24,25,26) up six places so that they are
before the two down-shot stacks (everything changes numbers but we
will ignore that)
it makes no difference to which set of feet appear, that is, no
difference to which Alpha mask is ignored.
with the sky stack before the two down-shot the images are like this:
Images 18, 19, 20 are bracketed exposure stack for the sky shot
Images 21, 22, 23 are bracketed exposure stack for one down shot
Images 24, 25, 26 are bracketed exposure stack for the other down
shot
I have tried adding and removing Alpha-Channel masks to the stack of
sky-shots and it makes no difference to the set of feet that appear
(or mask that is ignored)
So at first it seemed that whichever stack was at (18, 19, 20) was
the one whose Alpha mask would be ignored. But when the sky stack are
in those positions they do not get ignored - well actually it is
hard to tell about the sky - but I can be sure that they do not effect
which ground shot has it's mask ignored - it is always the first of
the ground shots that has its mask ignored no matter where that set of
shots is in the overall sequence and no matter which earlier shots
have or don't have masks.
I have now stitched this same panorama 20 times - and even the
computer is bored of dealing with the same images over and over :-)
and I think I can patch together a good result by combining two of the
stitches so I think I will now stop the quest to get one perfect
stitch.
I hope there is some useful debugging data there for someone ... feel
free to ask for more detail and or variations if it would help.
For example - I haven't tried switching to 8-bit TIFF would that be
worth a try?
all the best
George
On 16 Apr, 23:52, Bruno Postle <[email protected]> wrote:
> On Sat 11-Apr-2009 at 04:06 -0700, grow wrote:
>
>
>
> > So after all that long preamble my question is ... have other
> > people had problems with Alpha channel masking? Is there a known
> > error that makes it work on some projects and not on others? Is
> > there some fine detail that has to be "just so" with Alpha channel
> > masks?
>
> There is a bug in enblend that I haven't managed to characterise,
> where alpha masks in the first image in the sequence are discarded
> after the first blend. Try reorganising the photos using the Move
> up and Move down buttons in the Images tab before stitching.
>
> --
> Bruno
--~--~---------~--~----~------------~-------~--~----~
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
-~----------~----~----~----~------~----~------~--~---
