[CREATE] Lens correction database

sebastian sauer scb at lo-res.org
Tue Jan 8 11:15:53 PST 2008


hi,

[my last x-posting on this topic. reply-to set to
create at lists.freedesktop.org again; in the lack of a 
list dedicated to this topic.]

Tue 08 Jan 2008 16:23, Andrew Zabolotny wrote:
> From Tue, 8 Jan 2008 11:36:53 +0100
> sebastian sauer <scb at lo-res.org> wrote:
> 
> > basically all this lens+camera "defect" modelling is also very useful
> > in the geomatics/remote-sensing area. (of course also in computer
> > vision, medical imageing, etc.)
> I wasn't looking that far :) My guess is that specialised applications
> use cameras/lenses far from regular photographic equipment,
to make it less confusing, i'll make it simple :)
basically there are 2 widely used FOSS geomatics programs GRASS and OSSIM, 
and basically 3 different user-groups.


1/ people with really fat cash, having access to high-tech *modern* aerial
imaging equipment. (fuckin' expensive, really)

OSSIM supports this stuff..
https://svn.osgeo.org/ossim/trunk/ossim/src/apps/applanix2ogeom/
applanix2ogeom supports non-linear discrete radial lens distortion. 
basically you specifiy a parameterfile, which you get from the camera.


2/ people with still quite some budget, having access to old, classic
aerial photo-cameras.

..these cameras vary, but the higher-end modles among them are like that..
http://erosproject.cr.usgs.gov/osl/acspecs.html
http://erosproject.cr.usgs.gov/osl/reqtable.html
http://erosproject.cr.usgs.gov/osl/sixspecs.html
http://www.environment.sa.gov.au/mapland/camera_types.html

GRASS supports them, but AFAIK assumes that the images are rectilinear.
http://grass.itc.it/grass63/manuals/html63_user/photo.camera.html


3/ "ordinary" people that shoot with some standard of-the-shelf, over-the-counter 
DSLR, (often with wide-angle / fish-eye, in the hope to get a similar field-of-view 
as a professional aerial camera.)

these ppl. somehow not yet realize that exactly e.g. your software will help
them :)

e.g. this GRASS user, which for some reason decided to ignore me and my
comments..
http://lists.osgeo.org/pipermail/grass-user/2008-January/042487.html
http://lists.osgeo.org/pipermail/grass-user/2008-January/042496.html

.. see also there's a whole new generation comming, very cool..
http://openaerialmap.org/

and one of the god-fathers of FOSS geomatics Markus Neteler posted also
a very usful paper and a wiki-page..
http://lists.osgeo.org/pipermail/grass-user/2008-January/042494.html


> I'm wrong. In any case, I think lensfun will need some improvements to
> be useful in that area.
fact is, you guys here on this list are the actual experts on
lens-models. believe it or not :)

GRASS GIS has nothing for lens correction.

and OSSIM has decent models (what can compare well to commercial
geoformatics software) *but* nothing really advanced.
"OSSIM does have sensor models for aerial frame and pinhole cameras. 
A full blown camera model will have calibration information such 
as focal length and radial distortion parameters."
http://sourceforge.net/mailarchive/message.php?msg_name=0DDA504E-3A89-42EF-A81E-1EC4398C80B3%40mac.com


> > mmh, maybe an new mailing list, dedicated to photogammetry?
> If you can find more than two interested people, maybe it makes
> sense :)
agreed. maybe the Vision Workbench software makes it more clear how much
overlap we have on this topic..
http://ti.arc.nasa.gov/visionworkbench/

> But in any case my primary interest is photography, so I can't
> tell much on the subject you're interested in :)
sure, this also doesn't matter. as far as only lens / camera "defects" 
are our interests are the same.

also a side topic could be semi-automatic camera-calibration (that means
semi-automatic dedection of e.g. the lens parameters) as there is
already some decent software.
http://www.vision.caltech.edu/bouguetj/calib_doc/
http://graphics.cs.msu.ru/en/research/calibration/index.html
http://opencvlibrary.sourceforge.net/

..but this stuff could be added later to lensfun, as an alternative or 
enhancement to the lens database.


cheers,
s.


More information about the CREATE mailing list