[Insight-users] Fastest color-based segmentation ITK class

[Arturo Caissut]

Ok, now I better know my aim: the matter is that I have an Analyze image 
already "labelled" with a kind of K-means Classifier filter, and I need 
to provide a function to let the users extract all and only the voxels 
labelled with a certain intensity value (from 1 to 106 or so). Shall I 
use an Iterator or something? Maybe there's an ImageToImage family 
member which may help me?

Thanks,
Arturo



Luis Ibanez ha scritto:
>
> Hi Arturo,
>
> If you don't mind asking the user to provide seed points,
> you may want to consider the use of:
>
> http://www.itk.org/Insight/Doxygen/html/classitk_1_1VectorConfidenceConnectedImageFilter.html 
>
>
> It computes the mean and covariance matrix of a multi-component
> image around the seed points that you provide, and then uses
> a Mahalanobis distance in the parametric space of the image
> in order to select pixels with similar values.
>
>
> A second good option is to use SVD (singular value decomposition)
> in the color space.
>
> A third option is to run K-means.
> See examples in Insight/Examples/Statistics
>
>
>
>  Regards,
>
>
>    Luis
>
>
> -----------------------
> Arturo Caissut wrote:
>> Hi all,
>>
>> I'm looking for an ITK filter to extract all the voxels with a 
>> certain color from an Analyze file. Colored areas will be "condensed" 
>> in one or two islands, so I think it should be a pretty easy 
>> segmentation to do.
>> Could you suggest an ITK class to perform a fast (even if not so 
>> precise) segmentation?
>>
>> Thank you very much,
>> Arturo Caissut
>> _____________________________________
>> Powered by www.kitware.com
>>
>> Visit other Kitware open-source projects at
>> http://www.kitware.com/opensource/opensource.html
>>
>> Please keep messages on-topic and check the ITK FAQ at: 
>> http://www.itk.org/Wiki/ITK_FAQ
>>
>> Follow this link to subscribe/unsubscribe:
>> http://www.itk.org/mailman/listinfo/insight-users
>>
>