I think LensFun scales the coordinate system wrongly if there is a
calibration for cropfactor F1 but an image with cropfactor F2>F1.

 Malcolm Gillies has played a little bit with the Adobe models and
published this in this forum, and so I came to this conclusion.

 Imagine a distorted image of a checkerboard with cropfactor 1.  After
correction, you see the perfect checkerboard on your screen.  So far,
so good.

 Now, you double the cropfactor.  You simply use only the inner part
of the full-frame sensor.  LensFun corrects this image by dividing all
"r" coordinates by 2 and inserting this in the correction polynomial
f(r).  Note that LensFun normalises in a way that f(1) = 1.

 However, in the general case, this is wrong.  The reason is that
f(1/2) != 1/2.   So, you must find the scaling factor for which f(1/2
* scaling) = 1/2.  Adobe does this, LensFun does not (AFAIKS).  One
could create a test case by taking a badly distorted picture,
calibrate it, and crop it and see whether the calibration still works
(with the different cropfactor).  I cannot do that at the moment.

 Can anybody confirm or dispute that?

------------------------------------------------------------------------------
Put Bad Developers to Shame
Dominate Development with Jenkins Continuous Integration
Continuously Automate Build, Test & Deployment 
Start a new project now. Try Jenkins in the cloud.
http://p.sf.net/sfu/13600_Cloudbees
_______________________________________________
Lensfun-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/lensfun-users

Reply via email to