struggling with hugin vertical panorama and image rotation

[Eugeni Dodonov]

Hi,

I am struggling with Hugin while trying to create a vertical panorama
of these 2 images below. I don't know why (reflections on the image?
rotation? white balance? some stupid thing I am doing?), but Hugin
insists on automatically and incorrectly rotating or misplacing one of
the images, and obviously fails to create a valid panorama.

I know that it is perhaps a simple newbie question, but I have been
using Hugin for past few years - and now for the first time I feel
completely stuck. I simple don't know how to make it create the
panorama of those two images with it:
http://eugeni.dodonov.net/files/hugin/DSC03778.JPG
http://eugeni.dodonov.net/files/hugin/DSC03780.JPG

I tried using 'convert -rotate' and 'jhead -autorotate' before feeding
the images to Hugin, but without much success. I also manually added
all the control points manually, and set the 'roll' parameters on
Images tab to 0, but after I click "Align", the images get rotated or
completely misplaced. When going into Optimizer, I get extremely high
values for any settings or choose (or '0' if I choose Everything, but
then the resulting image gets completely insane).

Probably I am missing some simple step, so if any kind soul would be
able to show me what I am doing wrong I would be really grateful!

I am using hugin-2010.2.0-0.beta1 and autopano-sift-C-2.5.1.

[kfj]

On Sep 14, 2:48 am, Eugeni Dodonov <eugeni.dodo...@gmail.com> wrote:
> Hi,
>
> I am struggling with Hugin while trying to create a vertical panorama

...
I downloaded the images and my hugin processed them without any
problems, I didn't have to do anything to them previously, but maybe
the sending them over the internet did something to them? I noticed
there were no EXIF data about the hfov in the images, so I manually
entered a hfov of 50 degrees.

> I am using hugin-2010.2.0-0.beta1 and autopano-sift-C-2.5.1.

I'm using hugin 2010.1.0.5118 for Windows. CPs came out fine with
autopano-sift-c and with Panomatic.

[Eugeni Dodonov]

Interesting, I downloaded and installed Hugin 2010.0, and everything works perfectly. But with 2010.2.0 beta1 and current hg tip (4350:29beb35f0611) I cannot get a valid panorama no matter what I do.

Strangely enough, Hugin 2010.0 detects that there is no HFOV on images and asks me to input it (I usually use 50 degrees as well); 2010.2 and current hg version do not ask anything and just go ahead.

If there is anything I could do to help investigate this issue? Otherwise, I guess I'll stick with 2010.0 for now..

--
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 hugi...@googlegroups.com
To unsubscribe from this group, send email to hugin-ptx+...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/hugin-ptx

--
Eugeni Dodonov
http://eugeni.dodonov.net/

[kfj]

On Sep 14, 4:54 pm, Eugeni Dodonov <eug...@dodonov.net> wrote:
> Interesting, I downloaded and installed Hugin 2010.0, and everything works
> perfectly. But with 2010.2.0 beta1 and current hg tip (4350:29beb35f0611) I
> cannot get a valid panorama no matter what I do.
>
> Strangely enough, Hugin 2010.0 detects that there is no HFOV on images and
> asks me to input it (I usually use 50 degrees as well); 2010.2 and current
> hg version do not ask anything and just go ahead.
>
> If there is anything I could do to help investigate this issue? Otherwise, I
> guess I'll stick with 2010.0 for now..
>
>

1. sorry for being sloppy about the EXIF data. There are EXIF data in
your files, but they might not be enough. When I ran exiftool on them,
I found the focal length (5.8 mm) but no sensor data, and I think the
hfov can only be calculated if both the focal length and the sensor
size (and therefore the crop factor) are known (please correct me if
I'm wrong). If you are coming from rectilinear data and the output is
rectilinear as well, the hfov doesn't seem to matter so much as long
as it's some value near the mark, like 50 degrees here - and if you
have a 360 degree panorama the software can figure the fovs out. The
EXIF data also contain the rotation information - so my version of
hugin set the roll angle to 90 degrees straight away.

2. yeah, stick with whatever version works. I sometimes feel it's a
bit of a minefield - I download a promising version, use it for a
while and stick with it if it doesn't crash too often. I even
anticipate it crashing on certain operations and save the pto before I
perform them, because usually it can do it just fine and has just
messed up it's internal state so that the operation succeeds when you
reload the saved pto and try again. Since a lot of the functionality
is in separate executables, it also helps sometimes to replace some of
the helper applications with other versions.

3. I hope someone will look into the issue. I'm just a fellow user,
even though I sometimes feel like looking into the code myself. But
that feels like an uphill struggle (quote: compiling hugin yourself is
fairly extreme behaviour...) - and I'm on Windows, where things are
even more complicated...

with regards
KFJ

[Eugeni Dodonov]

On Tue, Sep 14, 2010 at 17:30, kfj <_k...@yahoo.com> wrote:

3. I hope someone will look into the issue. I'm just a fellow user,

even though I sometimes feel like looking into the code myself. But
that feels like an uphill struggle (quote: compiling hugin yourself is
fairly extreme behaviour...) - and I'm on Windows, where things are
even more complicated...

I'll try using 'hg bisect' on the code, maybe it will shed some light..

 

[Eugeni Dodonov]

Ok, so I bisected Hugin source to the commit which break processing of those pictures:

[eugeni@eugeni-x86_64 22:15:47 ~] $ cat /tmp/fix.bisect 

The first bad revision is:

changeset:   4071:b691961680b1

parent:      4062:a3988d6c4e7b

user:        tmodes

date:        Sun Jun 06 16:40:53 2010 +0200

summary:     Reset cropfactor to 1 and not to zero

The attached patch fixed the behavior for me - after applying it to current hg tip, the photos are aligned correctly. I don't know however why it is causing the problem, maybe any Hugin guru/hacker could take a look on it?

P.S.: I just updated Hugin hg to today's version, and looks like this patch was partly reverted today already.. I'll rebuild Hugin again to see if it works now.

[Eugeni Dodonov]

Yep, the latest version from mercurial fixes the issue for me. Thanks guys and sorry for the noise :).