[Insight-users] image segmentation ideas

[Luis Ibanez]

Hi Zein,

You may get rid of thing edges by applying mathematical
morphology operations in gray levels. In particular if
you want to eliminates membranes that appear brighter
than their surroundings you can use erosion followed
by dilation (equivalent to "opening").

GrayLevel mathematical morphology operations are implemented
in ITK by the filters:

http://www.itk.org/Insight/Doxygen/html/classitk_1_1GrayscaleDilateImageFilter.html
http://www.itk.org/Insight/Doxygen/html/classitk_1_1GrayscaleErodeImageFilter.html

They implement the more generic approach to morphomath,
where you can define structuring elements of arbitrary shape
and assign values to each one of the pixels in the structuring
element.

The default structuring element in these filters is the
traditional neighborhood of 1-pixel radius.

In principle the structuring element that you want to use
should have a radius equal to half the width of the widest
membrane you want to eliminate.

The erosion will eliminate the membranes and the subsequent
dilation will (hopefully) restore the area lost by other
bigger structures.

Please let us know if you have any dificulties using these
filters.


Thanks


Luis


================================================================


Zein Salah wrote:
> Hi friends,
> 
>  
> 
> I am trying to segment a 3D gray-level medical image. The area of my 
> interest is a relatively dark (somehow black)area that is surrounded by 
> a relatively lighter (gray and white) area. The problem with this region 
> is that there are a lot of thin gray edges (membranes) that are of no 
> importance for me, i.e. I do not need them. In other words, I want to 
> mark the whole region (and hopefilly, as if the thin edges that are 
> inside do not exit).
> 
>  
> 
> Can anybody give me ideas about what segmentation filters to use? hoe to 
> get rid of the thin edges without damaging the whole images.
> 
>  
> 
> I will be thankfull for any kind of help.
> 
>  
> 
> Zein
> 
>  
> 
> ->8<------------->8<------------->8<------------->8<------------->8<------------->8<-
> Zein I. Salah
> Universität Tübingen, WSI-GRIS
> Sand 14
> 72076 Tübingen
> Email: salah@gris.uni-tuebingen.de <mailto:salah@gris.uni-tuebingen.de>  
> / zeinsalah@hotmail.com <mailto:zeinsalah@hotmail.com>
> Tel.: (07071) 29 75465 (GRIS) , (07071) 96 44 39 (privat)
>