[ptx] Registration testcases

[alexandre jenny]

The idea of testcases is really good. 
I can provide many of them. Just have a look at my website :
http://www.kolor.com
There's actually 138 panoramas there and adding more today. This site is why
I'm not having too much time for sift code actually, but that's old story
now.

So I will provide some testcases : standard and fast one (small number of
pictures) and really complex and big one (with many exposures for example).

BTW : An idea for the future ;-) (I always have many idea. I post it here
just to keep a trace of it)
The next step after solving this SIFT features extraction, which seems to be
quite solved is to come back to nona and improve the blending for hdr
sampling. I explain :
Let's have 2 overlaping pictures (with some part overlap, not necessary the
whole picture). We assume that the matching algorithm already solved the
matrices and theses pictures are well located for stitch. Let's also assumes
we have the exif informations (Fnumber, exposure time). For each pixel, we
could with this information calculate the real pixel color by applying the
same calculation as in HDRshop and the output will directly by the pixel
color in float/double format achieving directly the full dynamic range. This
system could easily be extended to something many pictures with different
exposure for calculating one pixel in HDR.
By having that, you didn't need anymore to worry about overexposure,
underexposure, etc. Just take many pictures with different exposures and let
have the algorithm solves that. Moreover, blending is really simplified.
... It's just an idea for future ...

Now. Let's just create the keypoints algorithm for hugin ...

Bye
   Alexandre Jenny

Btw : Do you like my website ?

> I've put a few panoramas online, complete with source images.
> 
> http://wurm.wh-wurm.uni-ulm.de/~redman/gallery/pano_testcases
> 
> It contains a simple 4 pictures pano, a 24 pic half spherical 
> pano taken from the 161 m high church tower of Ulm (the town 
> where I live). Its really worth watching it with the native 
> PTViewer 0.4 or FSPViewer. :)
> 
> There is also one  pano with very similar structures, on 
> which the simple panosifter matching algorithm fails, and a 
> difficult single row 360°, with many moving people and 
> changing illumination. (there is one overlap where it is hard 
> to pick control points, even by hand...)
> 
> I hope we can use this set of images to compare different 
> registration techniques. If you have other panos that pose 
> difficulties not covered by the images here, please let me 
> know and I'll add them to the site.
> 
> Hmm, maybe the pure amount of images is a problem for some 
> implementations.. anyway, just use scale down the images, or 
> do partial panos then :)
> 
> Btw. I know that most of the panos will probably have been 
> shot more carefully than the images presented here.. luckily 
> we have the possibility to change stuff by hand :) (which was 
> not done for the panosifter tests on this gallery).
> 
> Btw. feel free to uses other programs on these panos as well, 
> I'd like to see how good or bad our results are.
> 
> ciao
>   Pablo
> -- 
> http://wurm.wohnheim.uni-ulm.de/~redman/
> Please use PGP
>