[Insight-users] Re: Watershed Filter for 3D data

[Gaëtan Lehmann]

Hi,

I didn't know the watershed can take hours to run. What is the size of  
your image ? Are you sure to have set the CMAKE_BUILD_TYPE to Release for  
your project ?
You may also be interested by the filters we have implemented with  
Reachard Beare  
http://insight-journal.org/InsightJournalManager/view_reviews.php?back=publications.php%3Fjournalid%3D4%26order%3D%26revision_display%3D&pubid=92
The tests we have run are showing an reasonnable gain of performance (with  
a 371x371x34 image, 7.47 sec for the ITK filter, and 4.61 with our filter,  
and even better if you want to perform a watershed from marker).
If you try it, please let me know the results you get. We haven't tested  
the effect of the image complexity, and I'm curious to know the behavior  
of our filter with a highly detailed image.

Gaetan

Le Thu, 27 Jul 2006 19:48:55 +0200, Josh Cates <cates at sci.utah.edu> a  
écrit:

> Hi Vijay,
>
> The watershed algorithm will take a very long time to process detailed  
> data.  The size of your dataset is less important.  Make sure you have  
> done the appropriate preprocessing of your data to remove noise and as  
> many uninteresting features as possible (for example, you might try an  
> anisotropic diffusion filter with conductance parameters set to preserve  
> the gradient edges of interest but blur other regions).  I also suggest  
> tuning parameters on a cropped region of your data before running on  
> your entire dataset.
>
> hope this helps,
>
> Josh.
>
> K Venkatraman, Vijay (GE, Research, non-ge) wrote:
>> Hi,
>>  I am using the watershed filter on  3D data, it seems to be taking a  
>> long time (couple of hours for 100slices)
>> The pipeline looks like: Analyze 3D data  
>> (*.img)->GradientMagnitudeImageFilter->WatershedFilter.
>>  The same pipeline works when the watershed filter is applied on 2D  
>> slices (extracted from 3D data), but I want them to work on the 3D data.
>>  I tried using lower threshold and level parameters.The data is little  
>> endian and unsigned short (is the filter dependent on endianness of the  
>> data). I have tried applying AnisotropicDiffusionImageFilter on the 3D  
>> data. It didn't help.
>>  Can anyone provide some help?
>>  Thanks
>> Vijay
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users



-- 
Utilisant le client e-mail révolutionnaire d'Opera :  
http://www.opera.com/mail/