[Insight-users] Re: Demo Pictures : Filling Holes in 3D

[Luis Ibanez]

Hi Jay,

Thanks for posting the screenshots of your volumes.

The holes that you are dealing with, seem to be a
good target for the Voting filters.

In particular for the HoleFillingVoting filter

(1)
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VotingBinaryHoleFillingImageFilter.html

        and its iterative version

(2)
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VotingBinaryIterativeHoleFillingImageFilter.html


You may want to try (1) first and see its effect
when you run it over an over your image. The filter
convert background pixels into foreground pixels
when they have a number of foreground neigbors over
a certain value. If that's effective, then you may
try to use (2).  Note that concavities are quite
similar to holes, as far as this filter is concerned.
Therefore you may also fill-in some of the indentation
in the anatomical structure.


These filters are N-Dimensional, so you don't have
any trouble applying them in 3D.

You will find examples in

              Insight/Examples/Filtering


You could also use the filter in a negated image in
order to get rid of the isolated noise that is around
the body (pretty much like a mathematical morphology
erosion. Note that Mathematical Morphology is actually
a particular case of voting filters).



Regards,


     Luis




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

> Hi, Luis,
>  
> Attached is a set of demo pictures I got from the screen shot.
....
snip
....
>  
> Demo1 is an OK scan, although we have many holes above the butt at the 
> back. Demo2 is a bad scan, since we lost too much information at front 
> of the waist. I'm thinking to use a 3-D approach to approximate the 
> waist region. Could you please give me any hint? Any opinions will be 
> highly appreciated. Thanks again. Have a nice day.
>  
> Regards,
> Jay
>