Wanted: Your distortion parameters

[Torsten Bronger]

Hallöchen!

Some of you will probably have .ini files (or something equivalent)
with the distortion data of your favourite panorama lenses. It
would be highly appreciated if you provided this data to the
http://lensfun.sourceforge.net/ project. At
<http://wilson.bronger.org/lensfun_coverage.html>, you can see the
lenses that we already have. But we have reliable FOV/focal length
data only for the Samyang 8mm f/2.8 and the Sony E fisheye
converter, so probably your data is welcomed anyway.

I'm the one who maintains Lensfun's lens database at the moment. So
if you have such data and are willing to give it to us, please send
an email to me containing:

- full lens model name
- camera model used for the pictures
- whether it was a partial or a 360° panorama
- the a, b, c parameters that Hugin generated
- projection that you used (rectilinear, fisheye, equisolidangle
etc)
- the FOV/focal length that Hugin generated

Or, equivalently, the ini file that should contain all this info.

Note that Lensfun doesn't use the d, e, g, t parameters of Hugin.
So if you have included them into the optimisation, Lensfun cannot
use the data. Possibly, it is feasible for you to load your old pto
file again and re-do the optimisation with all these parameters
fixed to zero.

Thank you very much!

Tschö,
Torsten.

--
Torsten Bronger Jabber ID: torsten...@jabber.rwth-aachen.de
or http://bronger-jmp.appspot.com

[Bruno Postle]

On Thu 10-Apr-2014 at 18:03 +0200, Torsten Bronger wrote:
>
>Some of you will probably have .ini files (or something equivalent)
>with the distortion data of your favourite panorama lenses. It
>would be highly appreciated if you provided this data to the
>http://lensfun.sourceforge.net/ project.

I still think Hugin (and the Calibrate Lens tool) could help with
this by automatically submitting optimised lens properties whenever
it finds a 'good stitch'. We had a discussion about this here:
http://thread.gmane.org/gmane.comp.misc.ptx/21536/focus=21604

The disadvantages are a lack of quality control and privacy issues.
Hugin would need to give a prominent option to not submit data, and
the quality isn't necessarily going to be worse than manually
created data.

The lens-submit proof of concept tool is still in the
Panotools::Script perl module. The collected database is here
(though it looks like nobody has submitted anything in three years):
http://oink.postle.net/lens.log

--
Bruno

[Terry Duell]

Hello Torsten,

On Fri, 11 Apr 2014 02:03:33 +1000, Torsten Bronger
<bro...@physik.rwth-aachen.de> wrote:

> Hallöchen!
>
> Some of you will probably have .ini files (or something equivalent)
> with the distortion data of your favourite panorama lenses. It
> would be highly appreciated if you provided this data to the
> http://lensfun.sourceforge.net/ project. At
> <http://wilson.bronger.org/lensfun_coverage.html>, you can see the
> lenses that we already have. But we have reliable FOV/focal length
> data only for the Samyang 8mm f/2.8 and the Sony E fisheye
> converter, so probably your data is welcomed anyway.

I have avoided saving lens calibration in a .ini file, instead opting to
do the calibration as per the Lensfun web site guidance, and then placing
my .xml files my local Lensfun database, and providing the file to the
Lensfun project.
I should point out that I use zoom lenses, and hence end up with
calibration data for a number of focal lengths.
Perhaps I misunderstand how all this works, but it seems to me that Hugin
can only do the calibration at one FL and hence you can end up with many
.ini files, none of which may match the FL for a particular project. It
just seemed a bit neater to have it all in the one file...maybe it doesn't
make any difference in the end.
I would also point out that that I provided a calibration for a Sigma
24-70 f2.8 IF EX DG HSM lens in March which hasn't made its way into the
list referenced above. I will send it on again, just in case it has gone
astray.

Cheers,
--
Regards,
Terry Duell

[Torsten Bronger]

Hallöchen!

Bruno Postle writes:

> [...]

>
> I still think Hugin (and the Calibrate Lens tool) could help with
> this by automatically submitting optimised lens properties
> whenever it finds a 'good stitch'. We had a discussion about this
> here:
> http://thread.gmane.org/gmane.comp.misc.ptx/21536/focus=21604

This may be an interesting "Wisdom of Crowds" approach in general
but not usable for the Lensfun database in my opinion. I consider
it a lot of work to convert the log file into a peer-reviewed
Lensfun database entry. The most prominent problem is the EXIF
hell. Besides, begging for contributions works quite well. We've
added 100 lenses in the last 6 months via
<http://wilson.bronger.org/calibration>.

By the way, while Hugin is the primary *source* of Lensfun data, I
don't really see much benefit for Hugin *using* Lensfun. Hugin
supports more fit parameters than Lensfun, the FOV always has to be
fitted anyway, and you want to have a profile of your lens specimen.
Moreover, Lensfun optimises the behaviour at the edges, while Hugin
optimises the behaviour at the stitching seams.

FWIW, I have never taken lens data from Lensfun when I stitched
panoramas, although I'm the most active Lensfun developer at the
moment.

Tschö,
Torsten.


P.S.: Funny enough, Robert Fendt states in the above referenced
discussion: "The current approach also has the problem that it sets
d=1-(a+b+c), which changes the apparent focal length of the lens."
Exactly about this we've had a discussion on this list this week.
;-)