[Insight-users] Re: [Insight-developers] strange background values andBinaryErodeImageFilter

[Gaetan Lehmann]

On Thu, 25 Aug 2005 14:04:45 +0200, Miller, James V (Research)  
<millerjv at crd.ge.com> wrote:

> I think your images look reasonable.  It might be nice if we
> can get the old binary erode to produce the same result.
>
> It will be great to have a FastIncremental erode capability.

It just need to be imported :-)
http://voxel.jouy.inra.fr/itkBinaryErodeImageFilter.h
http://voxel.jouy.inra.fr/itkBinaryErodeImageFilter.txx

Sorry, I have already renamed the class to BinaryErodeImageFilter for my  
own use, so you should have to rename it  
FastIncrementalBinaryErodeImageFilter (but I'm not sure it's useful, see  
below)

> Then we could replace the old dilate/erode with the FastIncremental
> (having the old routines subclass the new and have no real  
> implementation).

Why keep the 2 sets of classes (fast incremental ones and normal ones) ?
FastIncrementalBinaryDilateImageFilter is used only in  
SignedDanielssonDistanceMapImageFilter, and was not in the previous ITK  
release.

I think  FastIncrementalBinaryDilateImageFilter  should only be renamed  
BinaryDilateImageFilter and the current BinaryDilateImageFilter be dropped  
(and the same for binary erosion filter) :-)

>
> Jim
>
> -----Original Message-----
> From: insight-developers-bounces+millerjv=crd.ge.com at itk.org
> [mailto:insight-developers-bounces+millerjv=crd.ge.com at itk.org]On Behalf
> Of Gaetan Lehmann
> Sent: Thursday, August 25, 2005 7:42 AM
> To: insight-developers at itk.org; insight-users at itk.org
> Subject: [Insight-developers] strange background values
> andBinaryErodeImageFilter
>
>
> Hi,
>
> I have modified FastIncrementalBinaryDilateImageFilter to implement an
> efficient binary erosion filter.
>
> I'm not able to easily reproduce the way BinaryErodeImageFilter set the
> background value. In fact, it seems to be quite strange.
> I have attached 3 files :
>   + bw.png is the source image. There is 3 colors in it. The white is the
> foreground value for erosion while gray and black are
> both background.
>   + out-current.png is the result of a binary erosion with a circular
> kernel with a radius of 15
>   You can notice that the eroded disk in the middle of the shape is grey
> while the others eroded pixels are black. Also, the 2
> gray shapes in the top of the image are modified.
>   + out-fast.png is the result of the same erosion with the modified
> FastIncrementalBinaryDilateImageFilter. The white shape is
> the same (great !). All eroded pixels are black, shapes at the top of the
> image are not modified, and the gray disk is still
> visible inside the white shape.
>
> Can we use the second behavior for the binary erode image filter ? I find
> it really more consistent than the current behavior, and it should avoid
> to spend time to reproduce a quite strange behavior.
>
> Regards,
>
> Gaetan
>



-- 
Gaetan Lehmann <gaetan.lehmann at jouy.inra.fr>
Tel: +33 1 34 65 29 66
Biologie du Développement et de la Reproduction
INRA de Jouy-en-Josas (France)
Web: http://voxel.jouy.inra.fr