SURF

[Gabriel Landini]

The new jar for feature detection for ImageJ: http://labun.com/imagej-surf/
seems quite interesting (I wonder if this is related to the SIFT features?).
 
However despite the nice screenshots, there is no information at all on the
meaning of the results.
I guess the yellow lines indicate the size and orientation of a box
surrounding  the features, but what do the dot colours (red and blue) mean?
Thanks for any pointers.

Gabriel

[dscho]

Hi Gabriel,

On Wed, 2 Dec 2009, Gabriel Landini wrote:

> The new jar for feature detection for ImageJ:
> http://labun.com/imagej-surf/ seems quite interesting (I wonder if this
> is related to the SIFT features?).

The SURF algorithm is related to SIFT in that it tries to accomplish the
same goal -- extract robust landmarks in different scales -- but as far as
I understand, SURF is not patented.

Ciao,
Johannes

[Szilvi Mezey]

Gabriel, I'm guessing red is above threshold, blue is under threshold.

Szilvi Mezey

2009/12/2 Gabriel Landini <[hidden email]>

> The new jar for feature detection for ImageJ:
> http://labun.com/imagej-surf/
> seems quite interesting (I wonder if this is related to the SIFT
> features?).
>
> However despite the nice screenshots, there is no information at all on the
> meaning of the results.
> I guess the yellow lines indicate the size and orientation of a box
> surrounding  the features, but what do the dot colours (red and blue) mean?
> Thanks for any pointers.
>
> Gabriel
>
>

[Stephan Saalfeld]

Nope---it's the sign of the Hessian trace which is used to speed up
matching in SURF.

SURF is another scale invariant feature detection and description scheme
that is supposed to be simpler, faster and better than SIFT.  I am very
happy to see that implemented and, to play with it, wrapped
it into a Plugin similar to those for SIFT/MOPS.  Find that in an
experimental branch `ImageJ_SURF' of Fiji:

git checkout -b ImageJ_SURF origin/ImageJ_SURF
./Build.sh
./fiji

I wrote to the author and he will look at that and tell more during the
coming days.  I hope to see it integrated in Fiji soon :)

Best,
Stephan


On Thu, 2009-12-03 at 12:47 +0100, Szilvi Mezey wrote:

> Gabriel, I'm guessing red is above threshold, blue is under threshold.
>
> Szilvi Mezey
>
> 2009/12/2 Gabriel Landini <[hidden email]>
>
> > The new jar for feature detection for ImageJ:
> > http://labun.com/imagej-surf/
> > seems quite interesting (I wonder if this is related to the SIFT
> > features?).
> >
> > However despite the nice screenshots, there is no information at all on the
> > meaning of the results.
> > I guess the yellow lines indicate the size and orientation of a box
> > surrounding  the features, but what do the dot colours (red and blue) mean?
> > Thanks for any pointers.
> >
> > Gabriel
> >
> >

[Stephan Saalfeld]

Hi Xin,

it's still in that branch.   The author replied very friendly and is
fine with sharing through Fiji, but has no time for support.  Same
problem with me, so check the branch to have a look at SURF, it works
nicely---hopefully we'll find some time to merge it into a release
soon...

Best,
Stephan



On Wed, 2010-04-21 at 20:42 +0200, Xin ZHOU wrote:

> Hello, Stephan,
>
> I checked the latest version of Fiji
> SURF is not there.
>
> Any idea?
>
> cheers, Xin
>
> Stephan Saalfeld a écrit :
> > Nope---it's the sign of the Hessian trace which is used to speed up
> > matching in SURF.
> >
> > SURF is another scale invariant feature detection and description scheme
> > that is supposed to be simpler, faster and better than SIFT.  I am very
> > happy to see that implemented and, to play with it, wrapped
> > it into a Plugin similar to those for SIFT/MOPS.  Find that in an
> > experimental branch `ImageJ_SURF' of Fiji:
> >
> > git checkout -b ImageJ_SURF origin/ImageJ_SURF
> > ./Build.sh
> > ./fiji
> >
> > I wrote to the author and he will look at that and tell more during the
> > coming days.  I hope to see it integrated in Fiji soon :)
> >
> > Best,
> > Stephan
> >
> >
> > On Thu, 2009-12-03 at 12:47 +0100, Szilvi Mezey wrote:
> >  
> >> Gabriel, I'm guessing red is above threshold, blue is under threshold.
> >>
> >> Szilvi Mezey
> >>
> >> 2009/12/2 Gabriel Landini <[hidden email]>
> >>
> >>    
> >>> The new jar for feature detection for ImageJ:
> >>> http://labun.com/imagej-surf/
> >>> seems quite interesting (I wonder if this is related to the SIFT
> >>> features?).
> >>>
> >>> However despite the nice screenshots, there is no information at all on the
> >>> meaning of the results.
> >>> I guess the yellow lines indicate the size and orientation of a box
> >>> surrounding  the features, but what do the dot colours (red and blue) mean?
> >>> Thanks for any pointers.
> >>>
> >>> Gabriel
> >>>
> >>>
> >>>      
> >
> >  
>