[Insight-users] Questions about 3D Level Set Method

[Jay Li]

Hi, Luis,
 
Thanks a lot for your suggestion. Our radar can scan a person to get a 3-D cloud. Because of the shape of the body, such like big belly, big butt, some signal couldn't bound back to the receiver, and it's too expensivie for us to modify the hardware system. 
 
The application is to get circumferences of each slice along the z direction, so we performed 2-D ridge detection which is fast and sufficient for those slices which have good bounding back signal. For those slices with bad signals, the ridge contour is not closed. Actually we can only see a little bit information from 2-D slice view, which is not good enough, so we think about 3-D approach.
 
For the rendering part, we only plus and subtract a small number on z direction, corresponding to the spacing of z, of each ridge point to make 2 vetices of a triangle. the 3-D rendering is only for the developing purpose.
 
In my opinion, I don't know whether I'm right or wrong, 3-D level set or other deformable 3-D model could provide global information somehow, so holes filled by those methods might be more accurate. I'll try to get a screen view very soon. 
 
Thank you again. Have a nice day.
 
Regards,
Jay

Luis Ibanez <luis.ibanez at kitware.com> wrote:

Hi Jay,


What is your surface representation ?

I don't quite see how are you extracting a surface by doing
2D ridge detection along Z...

Is your surface a tube with the axis parallel to Z ?
in that case I would expect that you do 2D ridge detection
along XY and then connect ridges along Z.

Level Set are certainly a good option for filling holes,
however, depending on your surface representation, you
may find useful to use other simpler methods such as the
Voting filters:

http://www.itk.org/Insight/Doxygen/html/classitk_1_1VotingBinaryImageFilter.html
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VotingBinaryHoleFillingImageFilter.html
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VotingBinaryIterativeHoleFillingImageFilter.html


If your image looks more like an ultrasound (which probably
is to expect from a radar image...) you may consider some
model-based methods for extracting contours.

Maybe a screenshot of one of your images will help us to
understand the nature of your problem...


Regards,


Luis


-----------------
Jay Li wrote:

> Dear All,
> 
> Here I have some questions about ITK's level set classes. Our system is 
> a 3-D holographic radar. The image aquired is a 16-bit 256*256*224 
> "cloud" . We performed 2-D ridge detection of each slice along z axis 
> (1:224) to extract a "surface". Because of our radar's limitation, there 
> are a lot of holes and undesired points on the surface. Can level set 
> method smooth a 3-D surface? Can I input a volume data whose voxel 
> values are one at the ridge point and zero elsewhere as a target, and 
> use level set method to approximate a surface? Thanks.
> 
> Sincerely,
> 
> Jay
> 
> ------------------------------------------------------------------------
> Yahoo! Messenger
> Show us what our next emoticon should look like. Join the fun. 
> 
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users




		
---------------------------------
Yahoo! Messenger
 Show us what our next emoticon should look like. Join the fun.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20050404/ad824922/attachment.html