[CREATE] [hugin-ptx] Re: Lens correction database

[Pablo d'Angelo]

Hi Bernard,

Bernard Lang wrote:

>   I found this paper :
> http://groups.csail.mit.edu/graphics/classes/6.837/F99/devernay-faugeras_95b.ps.gz
> I realize that this paper is 10 years old.  Is the approach proposed
> in it obsolete or less accurate. It seems easier to use. I do not know
> whether there are other proposals of the kind.

I have thought about a similar approach, but I wasn't sure if it is reliable 
enough. The results in the paper look good, Maybe if I have some time to 
write the code, I'll try it.

> What is the importance of Image shift parameters "d" and "e" : Can
> they have an impact on the accuracy of other parameters ?  They have
> to be computed anyway for each camera, at least according to this
> tutorial : http://www.kekus.com/forum/showthread.php?t=31

Yes, they are usually dependent on the camera (play in lens mount etc.). 
However, for simple distortion correction, we can get away with assuming 
that they are located in the image centre. Due to differences between each 
camera, it is impossible to create a accurate (<1 of a pixel or so), generic 
distortion correction database anyway.

> 3 -
> 
> A database of the kind of EXIF data provided by each type of camera
> would also be quite useful.   Maybe it already exist, but I never
> saw it.  And technical documentation of vendors does not usually say
> anything about that.
> One advantage of such a database is that it would put pressure on
> camera makers to give accurate EXIF data.

exiftool contains GREAT support for all different kinds of metadata stored 
in images. However, this hasn't lead camera makers to be more open with
their EXIF extensions.

> 4 -
> 
> Why not include in the database the photos used for the calibration
> process, and the .pto file ? 

Actually, this is planned for dedicated calibrations.

> The only problem that I see is that some people might not want to have
> some of their pictures freely available in a free database.  This
> issue can be handled legally in a simple way : make the licence free
> and copyleft for calibration and analysis purposes, but reserve the
> right on the photos for any imaging purpose (unless explicitly stated
> in the comment fields of the picture, or otherwise).

Yes, for dedicated calibration images, they need to be available. However, I 
don't think this is a real problem, since the calibration shots will 
probably be dedicated images, without much artistic value.

ciao
   Pablo