stitcher and SIFT issues [Re: [ptx] faster stitcher]

[Pablo d'Angelo]

On Sun, 23 Nov 2003, alexandre jenny wrote:

> Hi pablo, 
> 
> It's cool your stitcher ! Let's compile it in windows, quick !
> BTW : Hint and Tips for windows compilation.
> Yes, you have to link against libjpeg, libtiff, libpng and fftw. I used some
> of the precompiled libraries available on sourceforge, but unfortunately, it
> doesn't worked (Some strange crash problem during premain _crt debug
> initialisation which crashed any vigra project before entering into _main
> function. But anyway, I solved it by recompiling first the libraries then
> vigra.

Ok, thats good news, then it should work on windows then.. I'm updating my
cross compilation enviroment to compile the stitcher there as well.

> I will try to compile your stitcher soon, to see it in action.

Please send me any project file you create.

> Another words on SIFT status, so good news and some bad news :
> Good news, I rewrote the sift detector with the use of vigra and it should
> work soon (I just made a quick and dirty matlab to cpp translation and
> didn't take time to debug the output, so It's still buggy but the whole
> algorithm is implemented).

Cool!

> The bad news : I asked for licencing issues and the sift detector is covered
> by a patent : Quote :
> "The patent covers the process used in SIFT, not the code itse;f.  Even if
> the code was rewritten, it would still be covered by the patent.  You could
> use the code you wrote to validate the scientific principles disclosed in
> Dr. Lowe's publications, but if you wanted to use Dr. Lowe's work for your
> benefit, commercial or otherwise, you would need a license." That's of
> course a bad news.

Not so cool. Well. do you know what exactly is patented? And in which
countries it is patented? Maybe we have to add a configure option depending
on the country. Maybe its also possible to work around the patent, if we
change or drop some part of the process.

Our application probably doesn't really need very robust scale invariance,
since the objects are approximately the same scale, in the usual project
(not considering minor effects as distortion introduced by the projection
etc.) It might also need some adjustment for non-rectilinear projections,
like fisheye.

ciao
  Pablo
--
http://wurm.wohnheim.uni-ulm.de/~redman/
Please use PGP