Hallöchen!
T. Modes writes:
> Am Montag, 18. November 2013 09:22:30 UTC+1 schrieb Torsten Bronger:
>
> [...]
>
> [...]
> In the documentation to both function it states
> <cite>
> The algorithm will look for a lens with crop factor not less than of given
> camera, since the mathematical models of the lens can be incorrect for
> sensor sizes larger than the one used for calibration.
> </cite>
>
> So the example you have given should not be possible according to
> the lensfun documentation.
Well, we have two issues here.
First, my example was erroneously swapped. I meant: Calibration was
made with APS-C (crop 1.5, aspect ratio 3:2), to-be-corrected image
was taken with Four-Thirds (2.0, 4:3).
Secondly, LensFun's documentation is wrong (even self-contradictory)
in the cited sentence. It was fixed to:
"The algorithm will look for a lens with crop factor not larger
than of given camera, since the mathematical models of the lens
can be incorrect for sensor sizes larger than the one used for
calibration."
> [...]
>
> So it would only be possible from calibration data from smaller
> crop factor cameras to deduce the calibration data for higher crop
> factor camera. But with libpano distortion model this is also not
> possible the other way around, because of the normalization. So
> you would need to recalculate all factors and not only to scale
> the factors.
I would have agreed till yesterday. I know that Adobe does more
than mere re-scaling of the "r" coordinate to get the sensor size
right, and even before this discussion here, I suspected a bug in
LensFun because of this.
However, LensFun behaves perfectly for cropped sensors, if the
calibration was done with a full-frame sensor. I can tell for sure
because LensFun now contains a testimage.py program which creates
distorted pictures for testing LensFun. The images are *exactly*
corrected in all cases. Since the distorting program works in the
coordinate system of the actual calibration sensor, cropping is
really simple here.
Besides, you can take a distorted picture, crop it manually in Gimp,
and let it be undistorted by LensFun with a camera with a
correspondingly higher crop factor -- works precisely.
I still have not understood why this works, though.
> [...]
>
> <snip>
>
>> 3. Introduce a new <aspect-ratio> tag in LensFun.
>
> One more point to consider: Even when you stay inside one crop
> factor, there are two possibilities to achieve a different aspect
> ratio: a) cropping height or b) a multi-aspect sensor. This have
> not been considered in the proposal.
The pair (crop factor, aspect ratio) must be correct. For
multi-aspect sensors, a virtual camera for each aspect ratio needs
to be defined in LensFun. The same is necessary for multi-crop
cameras (e.g. Nikon FX cameras in DX mode). The calling program
must pass the correct camera then.
We have now implemented the <aspect-ratio> tag, and a corresponding
attribute field in the lfLens class.
Tschö,
Torsten.
--
Torsten Bronger Jabber ID: [email protected]
or http://bronger-jmp.appspot.com
--
A list of frequently asked questions is available at:
http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to the Google Groups
"hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/hugin-ptx/878uvya55q.fsf%40physik.rwth-aachen.de.
For more options, visit https://groups.google.com/groups/opt_out.
