Hallöchen!

Matt Feifarek writes:

> On Fri, Jan 11, 2013 at 3:59 PM, Pascal de Bruijn
> <[email protected]>wrote:
>
>> To be honest, if this would be implemented, I wonder what would
>> happen to Lensfun. Contributions may drop even further, replacing
>> free distortion correction with a dependence on the grace of
>> Adobe.
>
> My hope is a converter from the .lcp files to the xml ones that
> lensfun needs, not to replace libraries.

I wrote such a converter for the distortion part.  My plan was to
fit the Adobe curve with the LensFun polynomial.  It didn't work.  I
haven't yet had a close look at the mathematics involved (haven't
had time for it anymore for the last three months), but my
preliminary analysis was that it cannot work: You cannot fit one
polynomial with a different one reasonably.  Maybe because both are
orthogonal?  Anyway ...

If this turns out to be true, the only way to have a converter is to
embed Adobe's models into LensFun.  Actually, given the fact that
LensFun implements already many models, I don't think this would be
very difficult.  But I don't have time for it.

>> Any particular problem with generating the corrections?
>
> It's been in finding good images, yes. I've found some contrasty,
> rectilinear architecture, but without correction, the images have
> lots of chromatic abberation... which makes it hard to find
> pin-prick accuracy for the control points.
>
> The whole hugin process kindof sucks, too.

Have you had a look at
<http://wilson.homeunix.com/lens_calibration_tutorial/>?

>> Typically a lot of the problem is bad source imagery. I had some
>> straight lines printed on A2 sized piece of paper and stuck it to
>> a sheet of foam-board to keep it perfectly straight.
>
> I'll try that, good idea. Architecture is unreliable!

Modern architecture is okay.  The PTLens library relies on it, and
its author does it for a living after all.  If you want to be
absolutely sure, use string+plummet, and take three pictures which
you calibrate in one run: One with the string close to the long
border, one to the short border, and one parallel to the long border
but running closer to the centre of the image.

The problem is that distortion parameters may depend on focus
distance.  I have no study at hand about how frequent and how
serious this problem is but I recommend taking pictures at a minimum
distance of 8 metres.  For example,
http://en.wikipedia.org/wiki/Breathing_(lens) magnifies the image
which changes the a, b, c parameters because their coordinate system
is magnified, too.  See also
http://www.dxo.com/en/photo/dxo_optics_pro/features/optics_geometry_corrections/distortion#ancre_how_complex
point "3. Focus dependence".

For wide-angles and fisheyes, night sky calibration may be an
option, see <http://www.johnhpanos.com/starcal.htm>.  This morning,
I took pictures for this, but I haven't yet given the calibration a
try.

Tschö,
Torsten.

-- 
Torsten Bronger    Jabber ID: [email protected]
                                  or http://bronger-jmp.appspot.com


------------------------------------------------------------------------------
Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS,
MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current
with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft
MVPs and experts. SALE $99.99 this month only -- learn more at:
http://p.sf.net/sfu/learnmore_122912
_______________________________________________
Darktable-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-users

Reply via email to