Test builds
Stephan Hegel
This morning I've found the time to build the following libs / programs
on my OpenSuSE 10.3, x86_64 box:
autopano-sift-C.r2629
enblend.20080113092448
hugin.r2629
libpano13.r787
Everything built fine.
I ran Yuv's hugin_6aroundtilted_testcase.pto through hugin_stitch_project.
Also almost fine: it produces a panorama tif file, but the command doesn't
return to the command line, even after closing the output window. When
trying to open the resulting tif file in xv, it warns about an unknown xml
tag.
Two questions arose:
1. Enblend is still in CVS, right ?
2. In my pano directory I've found versions of clens, freepv, panoglview
and Panotools-Script. What's the current status of this programs ?
Thanks & Rgds,
Stephan.
Yuval Levy
Stephan Hegel wrote:
> autopano-sift-C.r2629
> enblend.20080113092448
> hugin.r2629
> libpano13.r787
built exactly the same on my 32bit ubuntu 7.10 box.
> Everything built fine.
same here.
> I ran Yuv's hugin_6aroundtilted_testcase.pto through hugin_stitch_project.
did you try to start a project from scratch, i.e. use the assistant to
load the images?
when I do so, I still get the Error 255 message while trying to execute
autopano-sift-C. I get the same Error 255 when I try to run autopano
from the Images tab.
EXIF data is read, but something along the chain goes wrong and so the
FOV reading and lens type are not correct. I filed a bug report at
<http://sourceforge.net/tracker/index.php?func=detail&aid=1870369&group_id=77506&atid=550441>
When I use exiftool to read the EXIF data from the CR2, it reports a FoV
of 110.4 degrees and a focal lens of 8.0mm with a scale factor of 1.6,
which is 12.5mm in 35mm equivalent.
When I use exiftool to read the EXIF data from the Lightroom converted
TIF, it reports a FoV of 108.4 degrees and an 8.0mm focal lens which it
says it is 13mm in 35mm equivalent. I guess the small changes are part
of the conversion process.
hugin sets the lens type to rectilinear and hfov to 85.5 degrees. This
would be correct if the lens was rectilinear, but in this case it is not.
My guess is that hugin calculates the hfov based on the focal distance
and the assumption that the lens is rectilinear.
Since both FOV and focal distance are available in EXIF, would it be
possible to make the above calculations for different type of lenses and
set the lens type based on the one that gets the closest result to the
EXIF reported FOV?
Thanks
Yuv
Pablo d'Angelo
Yuval Levy schrieb:
> Hi Stephan,
>
> Stephan Hegel wrote:
>> autopano-sift-C.r2629
>> enblend.20080113092448
>> hugin.r2629
>> libpano13.r787
>
> built exactly the same on my 32bit ubuntu 7.10 box.
>
>
>> Everything built fine.
>
> same here.
>
>
>> I ran Yuv's hugin_6aroundtilted_testcase.pto through hugin_stitch_project.
>
> did you try to start a project from scratch, i.e. use the assistant to
> load the images?
>
> when I do so, I still get the Error 255 message while trying to execute
> autopano-sift-C. I get the same Error 255 when I try to run autopano
> from the Images tab.
>
> EXIF data is read, but something along the chain goes wrong and so the
> FOV reading and lens type are not correct. I filed a bug report at
> <http://sourceforge.net/tracker/index.php?func=detail&aid=1870369&group_id=77506&atid=550441>
As I wrote, there is no standard exif tag that encodes the lens type or the FOV.
> When I use exiftool to read the EXIF data from the CR2, it reports a FoV
> of 110.4 degrees and a focal lens of 8.0mm with a scale factor of 1.6,
> which is 12.5mm in 35mm equivalent.
Exivtool is the most advanced program for reading the image metadata (not
just EXIF tags) and it knows TONS of custom fields and probably works around
lots of inconsistencies in the metadata. Unfortunately, it is written in
perl and cannot be easily used from within hugin.
One way would probably be to extract the metadata, convert it to a standart
format such as xmp (if xmp can hold the FOV) and then read that with exiv2.
exiv2 itself can also read some MarkerNodes and RAW file meta information,
but I'm not sure if it can access the tags that exiftool uses to determine
the FOV.
If somebody can find out how exiftool computes the FOV, then it might be
possible to use the same approach with exiv2 (maybe..).
> My guess is that hugin calculates the hfov based on the focal distance
> and the assumption that the lens is rectilinear.
Exactly.
> Since both FOV and focal distance are available in EXIF,
As said above, unfortunately the FOV is not stored in standard EXIF data.
> would it be
> possible to make the above calculations for different type of lenses and
> set the lens type based on the one that gets the closest result to the
> EXIF reported FOV?
Yes that would be nice, unfortunately I don't have the time to dig into this
myself.
ciao
Pablo
Bruno Postle
>
> 1. Enblend is still in CVS, right ?
Yes.
> 2. In my pano directory I've found versions of clens, freepv, panoglview
> and Panotools-Script. What's the current status of this programs ?
clens is now superceded by fulla which is part of hugin.
freepv needs some small amount of work to get going:
a. The trunk isn't the latest code, you need to build
branches/branch_leonox
b. The autoconf build system doesn't work and needs to be removed,
use the cmake build system.
c. You need this patch to actually view QTVR:
http://sourceforge.net/tracker/index.php?func=detail&aid=1811163&group_id=148373&atid=772024
d. You may find you need one or other of these fixes:
http://sourceforge.net/tracker/index.php?func=detail&aid=1832092&group_id=148373&atid=772024
http://sourceforge.net/tracker/index.php?func=detail&aid=1811160&group_id=148373&atid=772024
http://sourceforge.net/tracker/index.php?func=detail&aid=1809816&group_id=148373&atid=772024
http://sourceforge.net/tracker/index.php?func=detail&aid=1809812&group_id=148373&atid=772024
panoglview is/should be superceded by freepv.
Panotools-Script is alive and well and due for a new 0.12 release.
This is basically a perl library for reading, writing and
manipulating hugin .pto project files (yes, the name isn't so
great).
--
Bruno
Gerry Patterson
I have time to work with Yuv on this. I have recently switched to autopano-c and I have enough perl knowledge to see what is going on inside of exiftool.
First, I need to duplicate the error 255 that Yuv is seeing. Is there a place I can get the project file and source images from?
- Gerry
Yuv
On Jan 13, 2:48 pm, "Gerry Patterson" <thedeepvo...@gmail.com> wrote:
> I have time to work with Yuv on this. I have recently switched to
> autopano-c and I have enough perl knowledge to see what is going on inside
> of exiftool.
For the exif issue, after a quick analysis my understanding is that
the lack of standard mention by Pablo is the quagmire and he is right
to better steer hugin off that quagmire and only take information that
is passed from the metadata reading library.
The author of ExifTool has done a lot of work to get to the results
that display on my screen and that I unknowingly assumed to be part of
the standard.
I believe the issue should be solved at its root, and I posted to the
Exiv2 mailing list:
<http://uk.groups.yahoo.com/group/exiv2/message/1014>
> First, I need to duplicate the error 255 that Yuv is seeing. Is there a
> place I can get the project file and source images from?
you can get the full hugin project, the six RAW files and the six
Lightroom-converted TIFF files (it's a 173MB download).
I am not sure if the error 255 is on my side. At the next possible
reboot I will see if I can reproduce the issue on my AMD64 box, and I
will try to "downgrade" autopano-sift-C and the libpano13 to the last
revision for which I had autopano-sift-C working properly (I am afraid
I did not note down which version it was, but I will find out).
Thanks for your effort, let me know if I can be of any help.
Yuv
Gerry Patterson
Hi Gerry,
Hello,
I am on the exiv2 list as well and read your post. I happen to be also using canon gear so I have access to a lot of RAW CR2 files. In the bug report you posted to Hugin's tracker, you mentioned you could see FOV being reported from exiftool on CR2 files. I am not seeing this. I have been known to miss things starting me in the face, but I checked twice, honest.
I am downloading the test case to see what is happening with the error on autopano-c.
Best Regards,
- Gerry
Yuval Levy
> In the bug report you posted to Hugin's tracker, you mentioned you could
> see FOV being reported from exiftool on CR2 files. I am not seeing this.
> I have been known to miss things starting me in the face, but I checked
> twice, honest.
don't worry. I too, despite laser-corrected 20/20 vision, miss things
starting me in the face.
when I drag a CR2 file on the latest exiftools that I downloaded two
days ago, I find a line stating:
Field Of View : 110.4 deg
It is the sixth line from the bottom of the output not including the
"press any key" output, just above Focal Length and below Circle of
Confusion.
Yuv
Pablo d'Angelo
> Gerry Patterson wrote:
>> In the bug report you posted to Hugin's tracker, you mentioned you could
> > see FOV being reported from exiftool on CR2 files. I am not seeing this.
> > I have been known to miss things starting me in the face, but I checked
>> twice, honest.
>
> don't worry. I too, despite laser-corrected 20/20 vision, miss things
> starting me in the face.
>
> when I drag a CR2 file on the latest exiftools that I downloaded two
> days ago, I find a line stating:
>
> Field Of View : 110.4 deg
exiftool on my system (v 6.90) does not output the Field of View of your
test files either, so this must be a new feature of release 7.10 or so.
ciao
Pablo
Gerry Patterson
Hello,
I was able to download the test case. However, when I load the project, clear out the points and press the align button I don't get an error. So I closed hugin and started from scratch (not loading the project). I used the assistant tab to load the images. The autopano output window opened and it completed normaly, no error. I am running (K)ubuntu 7.10 32 bit. Perhaps you can run the autopano command from a terminal window and check the exit status (echo $?). If you get 255. then something is going on inside autopano.
I updated my version of exiftool and now see the FOV tag. I looked at the code in exiftool and the computed FOV is only valid for Rectilinear images..... Hugin changes its FOV calculation for fisheye images. I'll keep looking at this to try to understand more of what is going on. But, in short, I wouldn't trust the output of the FOV field in exiftool for your fisheye images.
Best Regards,
- Gerry
Yuval Levy
> exiftool on my system (v 6.90) does not output the Field of View of your
> test files either, so this must be a new feature of release 7.10 or so.
yes, I am using release 7.10.
Yuv
Yuval Levy
Gerry Patterson wrote:
> Perhaps you can run the autopano command from a terminal window
> and check the exit status (echo $?). If you get 255. then something is
> going on inside autopano.
I'll check at the next possible opportunity and will report back.
> I updated my version of exiftool and now see the FOV tag. I looked at
> the code in exiftool and the computed FOV is only valid for Rectilinear
> images..... Hugin changes its FOV calculation for fisheye images. I'll
> keep looking at this to try to understand more of what is going on.
> But, in short, I wouldn't trust the output of the FOV field in exiftool
> for your fisheye images.
bizarre - the number I got from exiftool for my fisheye images was
correct, and it was the same number hugin gave me once i manually
changed lens type from rectilinear to fisheye. i have not tested with
rectilinear images.
Yuv
Gerry Patterson
Hello,
I was wondering, how you know what the correct value of the FOV is for your lens? I found this site that computes FOV for various configurations.
http://www.frankvanderpol.nl/fov_pan_calc.htm
I did notice the assumption of rectilinear lens type produces the wrong degrees of view. To correct this one needs to change the setting and then get hugin to recompute the value. I just re-entered in the focal length and pressed ENTER. Perhaps one could add an event to the dropdown to convince hugin to recompute the degrees of view when the lens type changes. If this sounds like a good idea, let me know and I will try it.
Side note:
I have never used a fisheye lens. It took me a while to figure out why your pictures seemed to be at an angle. I am guessing you are keeping the vertical FOV at 180 and rotating as much as you can to increase the horizontal FOV. Basically, rotating the hypotenuse of the sensor to vertical. Which pano head are your using to accomplish this?
Best Regards,
- Gerry
Yuval Levy
Gerry Patterson wrote:
> I was wondering, how you know what the correct value of the FOV is for
> your lens? I found this site that computes FOV for various
> configurations.
>
> http://www.frankvanderpol.nl/fov_pan_calc.htm
I go by a mix of calculation, intuition, approximation. The Sigma
fisheye is approximately 181 degrees of circle on 24mm diameter. The
sensor of my 350D is about 22mm wide and has a 3:2 form factor. So
vertically I have an estimated 166 degrees (92%) in portrait mode, which
means somewhere around 111 horizontally, which is pretty close to
exiftools' result and to hugin's result when I set the lens type to
fisheye. So I assume both programs make the same calculation since when
I feed them with the same data they return the same, correct answer.
> Perhaps one could add an event to the
> dropdown to convince hugin to recompute the degrees of view when the
> lens type changes. If this sounds like a good idea, let me know and I
> will try it.
I find it an excellent idea.
> Side note:
> I have never used a fisheye lens. It took me a while to figure out why
> your pictures seemed to be at an angle. I am guessing you are keeping
> the vertical FOV at 180 and rotating as much as you can to increase the
> horizontal FOV. Basically, rotating the hypotenuse of the sensor to
> vertical. Which pano head are your using to accomplish this?
yes, your guess is right. moreover, Canon sensors are slightly smaller
than Nikons, so no full 180° on the vertical (you will notice that the
short sides of the sensor touches the circle of the image).
AFAIK the first to publish the idea of tilting the sensor was Helmut
Dersch on his old website. You can see my self-modded pano head at
<http://www.photopla.net/wwp0506/newnpa.php> - the commercial design by
Agnos is the only one known to me to use this orientation and it was too
clumpy and heavy for my taste.
I like this kind of workflow because it minimizes shooting time and
maximises the use of the sensor surface. Especially the time
minimization is particularly important when walking / touring a place
with non-photographers whose patience is being tested at every sphere.
Though sometimes even the least patient of the non-photographers can
wait for at least five of the six frames :-)
<http://www.photopla.net/071224foretperdue/> (first four are for a time
lapse, the other two are variations on the second theme).
Yuv (on late night digression. got Visual Studio 8 Express Edition
installed, thanks, Pablo, for the SDK, will try that sometimes next week)