Posted by
Stephan Saalfeld on
Oct 21, 2008; 11:14am
URL: http://imagej.273.s1.nabble.com/ImageJ-SIFT-and-MOPS-tp3694680p3694683.html
Dear Xin,
The DoG detector detects blobs which includes junctions. This regards
the bare image intensities only and might totally differ from what you
expect to be a junction or blob. I doubt that the DoG detector is an
appropriate thing for detecting fractures in bones (I assume you're
talking about X-ray images?). Actually, I don't see how DoG detections
or SIFT features would help you in this field, but may be this is
because I have totally no idea about how your images look like.
Assuming that it is X-ray and that you know something about the angle of
projection and the imaged bones it could eventually help to have a look
at scale-invariant edge and ridge detection as described by Lindeberg
ftp://ftp.nada.kth.se/CVAP/reports/cvap191.pdf
Sorry for not being able to deliver a fast "do-it-so" answer.
Best regards,
Stephan
On Mon, 2008-10-20 at 21:50 +0200, Xin ZHOU wrote:
> Hello, Stephan,
>
> I'm using SIFT plugin.
>
> I've got a question concerning about the SIFT plugin.
> Actually I'm working on Fracture images and my main goal is to detect
> fracture in bone.
> My problem is that the points found don't correspond exactly to what I
> want, which is : edges and where there's high changes in gradients.
>
> I've changed the "MAX_CURVATURE" in FloatArray2DScaleOctaveDoGDetector.java.
> I increased from 10 to 100. It works better but not perfect.
>
> Do you have any idea how to better detect those edge points as well?
>
> cheers, Xin
>
> Stephan Saalfeld a écrit :
> > Dear Azizisya,
> >
> > the source code is GPL with the exception of SIFT since the UBC holds a
> > patent on this. You find it in the Fiji git-repositories:
> >
> >
http://pacific.mpi-cbg.de/cgi-bin/gitweb.cgi?p=mpicbg.git;a=summary> >
> > I am curious what you have in mind doing with it.
> >
> > Best regards,
> > Stephan
> >
> >
> > On Mon, 2008-10-20 at 04:41 -0700, azizia abdullah wrote:
> >
> >> Dear All,
> >>
> >> I am using imageJ to analyze natural images. In this case, I am using
> >> SIFT and MOPS
> >> (
http://pacific.mpi-cbg.de/wiki/index.php/Feature_Extraction) packages
> >> for automatic feature extraction and object detection. Both of them
> >> are based on finding salient regions of points in an image. Does
> >> anybody knows where I can download the SIFT and MOPS source codes. Or
> >> does anybody has implemented these techniques on a dense grid of
> >> uniformly spaced cells or overlapping grids.
> >>
> >> best regards
> >> azizisya.
> >>
> >> __________________________________________________
> >> Do You Yahoo!?
> >> Tired of spam? Yahoo! Mail has the best spam protection around
> >>
http://mail.yahoo.com
> >>
> >
> >
>