AW: [Insight-users] Question about Conntected Regions

[Luis Ibanez]

Hi Zein,

I agree with you in that 30 to 40 pixels is too
wide to apply matematical morphology for elimina-
ting the connecting region.

This situation lead to think that you may want
to solve better your segmenation stage. Instead
of trying to fix the resulting binary image.

Look for any differences in the pre-segmented
image between the region you want to conserve
and the region you want to eliminate. It looks
like an additional segmentation criterion will
help you to arrive to a binary image holding
only the region you consider interesting.


Regards,


    Luis


-----------------------------------------
salah wrote:
> Hi Luis, Thanks for the answer!
> 
> Erosion-Dilation would work fine if the connecting-region is fairly narrow. 
> I roughly draw the attached image just to ilustrate the idea. In my 
> image, this tunnel-like region is probably 30-40 voxel wide!!! 
> 
> If I use erosion-dilation, I must use a big radius for the structuring element. That
> means:
> 1. The remaining object would be extremely excessively affected, and 
> 2. It would take a long time. Considering my images of size 250x220x90, the erosion 
> filter alone requres about 75 seconds for radius 2, about 200  seconds for radius 3.
> 
> you could imagine how long it will take for radius 20 !!!
> 
> Thanks however for the suggestion. 
> 
> Zein
> 
> -----Ursprüngliche Nachricht-----
> Von: Luis Ibanez [mailto:luis.ibanez@kitware.com]
> Gesendet: Donnerstag, 20. März 2003 20:49
> An: salah
> Cc: ITK Users
> Betreff: Re: [Insight-users] Question about Conntected Regions
> 
> 
> Hi Zein,
> 
> This looks like a classical target for
> mathematical morphology.
> 
> You may disconnect the two regions using
> 'openning' which is simply an erosion
> followed by a dilation.
> 
> You will have to use a circular structuring
> element with a diameter equal to the width
> of the structure that you want to eliminate.
> 
> Note that as a collateral effect of the
> opening, the final shape will be more rounded.
> 
> 
>     Luis
> 
> 
> -----------------------------------------
> 
> 
> salah wrote:
> 
>>Hi everybody,
>>
>>I have a binary image which is a result of a segmentation process. As usual, it has 
>>a large connected object. This object contains two big regions that are connected
>>at some relatively narrow tunnel-like region. The attached figure illustrates the 
>>idea in 2D. 
>>
>>I need to disconnect these two regions as much automatically as possible. Actuallty, 
>>I want to remove one one of them without affecting the one that I want to keep. 
>>
>>Could anybody please help me doing that. I would be very thankfull for any kind of 
>>help, ideas, suggestions, ...
>>
>>
>>Zein
>>
>>->8<------------->8<------------->8<------------->8<------------->8<------------->8<-
>>Zein I. Salah 
>>University of Tübingen, WSI-GRIS, Sand 14, 72076 Tübingen 
>>Email: salah@gris.uni-tuebingen.de
>>Tel.: (07071) 29 75465 (GRIS),           Fax: (07071) 29 54 66
>> <<imagedisconnect.jpg>> 
>>
>>
>>------------------------------------------------------------------------
>>
> 
> 
> 
> 
>