[ptx] SIFT algorithm question

[Sebastian Nowozin]

Hi Alexandre, hi list :)


On Fri, Jan 02, 2004 at 02:54:27PM +0100, alexandre jenny wrote:

> Yes, that's right. I also get the same obversation, but as you know, people
> are always hiding thinks when publishing algorithm. So when you look at the
> last paper of lowe called : "distinctive image features from scale-invariant
> keypoints", you can see that the DoG stage (difference of gaussian) is a
> little bit different, and it tells much more for implementing this stage.
> This paper has also many new improvement for the averall algorithm.
 
Ah, great! Just the answer I needed :-) (I hope in his newer paper there are no
hidden tricks :-) But the more I read the first paper today I lost fear of all
the mathematics and I think I could do an implementation in the next week or
two.

Btw, I tried later with a few larger pictures (2 and 5MP) and there the effect
is greatly reduced and more fine lines are visible. I already wondered why the
size/sigma of the gaussian filter remains constant through the pyramid. Maybe
the paper you mentioned explains a bit more.


> Anyway, here is the source code for the sift extractor I rewrote in matlab:
> http://www.le-geo.com/temp/sift6.m
> It seems to work but I didn't get time to check the last stage of the
> algorithm which is the keypoint descriptor construction.
 
Cool, I will definitly check that out :)


> BTW : one think could be very cool : Dr Lowe released a binary linux version
> of the keypoint detector here. If someone have linux (I didn't), it could be
> cool to compare the result from hi software and from my to see if I'm not
> doing it wrong : http://www.cs.ubc.ca/~lowe/keypoints/
 
Ah, at least with the included images it runs impressively. If only half of the
accuracy/speed is reached in our code, automatic panorama stitching should be
easy :-)


> Bye
>   Alexandre

Thanks for the quick and interesting answer,
Sebastian

-- 
nowozin at cs.tu-berlin.de --- http://user.cs.tu-berlin.de/~nowozin/