[Insight-developers] Refactoring SimpleFuzzyConnectedness

[Yinpeng Jin]

Hi, Luis:
I agreed with your suggestion.
The only thing need to discuss is that the AffinityMap was defined with
value (0,1), while were stored as unsigned short with value (0,65535) in the
implementation, as suggested by Jay's original paper.
The other concern is that by breaking it down, itself will not be a
segmentation filter per se...
I don't know if Jay or his group has anything to say...

Yinpeng.



----- Original Message ----- 
From: "Luis Ibanez" <luis.ibanez at kitware.com>
To: "Yinpeng Jin" <yj76 at columbia.edu>; "Insight Developers List"
<insight-developers at public.kitware.com>
Sent: Tuesday, May 18, 2004 6:48 PM
Subject: [Insight-developers] Refactoring SimpleFuzzyConnectedness


>
> Yin,
>
> We have been looking closely at the FuzzyConnecteness
> filter and found that it can produce nice segmentations
> in many cases.
>
> However its use is clumsy because the output is the
> binarization of the AffinityMap.  The filter is itself
> computing the thresholding of the affinity map and
> providing this as final output.
>
> Since most of the time goes into the computation of the
> affinity map, it is a waste of time to play with the
> thresholding value, because the filter in the pipeline
> recomputes the affinity map every time.
>
> There is an UpdateThreshold() that can be called
> in order to recompute only the thresholding but
> unfortunately this doesn't adheres to the architecture
> of the data pipeline. The recomputations should go
> only through the Update() method.
>
> It looks like this filter will be much easier to use
> if the official output is the AffinityMap and not its
> binarization.
>
> Users can always plug this output into the threshold
> filter or a threshold ImageAdaptor.
>
>
> If the output is the Affinity map, it is so much easier
> to figure out the right level to binarize the image in
> order to get a reasonable segmentation.
>
>
>
>
> Please let us know if you have any objection to these
> modifications.
>
>
>
>    Thanks
>
>
>
>       Luis
>
>
>
> ---------
>
>
> _______________________________________________
> Insight-developers mailing list
> Insight-developers at itk.org
> http://www.itk.org/mailman/listinfo/insight-developers