[Insight-users] watershed with MRI

[Joshua Cates]

Hi Lydia,

The watershed algorithm was intended to be used with real-valued data
types (i.e. floats or doubles).  It will not work directly on multiple
component data such as RGB.  For best results, you should cast your
unsigned char image to floating point.  

Note that you will usually not pass the image you are trying to segment
directly into the watershed filter.  Typically some preprocessing is
required.  The watershed algorithm treats its input as a height function
where high values represent "ridges" or edge features in the image.  For 
RGB data, for example, you need to extract scalar image features from that 
data (i.e. gradient magnitude, curvature) first.

Please see the documentation in 

http://www.itk.org/Insight/Doxygen/html/classitk_1_1WatershedImageFilter.html#_details

and

http://www.itk.org/HTML/WatershedSegmentationExample.html

There is a more exhaustive description of the algorithm in the 
InsightDocuments repository (see http://www.itk.org/HTML/Download.php to 
download) under InsightDocuments/CourseWare/Training/Tutorial.pdf.  Also 
under 
InsightDocuments/Validation/WatershedSegmentation/WatershedSegmentation.pdf.

Hope this helps,

Josh.



______________________________
 Josh Cates			
 School of Computer Science	
 University of Utah
 Email: cates@sci.utah.edu
 Phone: (801) 587-7697
 URL:   http://www.sci.utah.edu/~cates


On Wed, 8 Jan 2003, lydia coin wrote:

> 
> Did anyone succeed to use the watershed algorithm with MRI images with a pixel type of unsigned char. I failed to make it work and I'm wondering if it works only with RBG images
> 
> thanks for any hints
> 
> Lydia
> 
> 
> 
> ---------------------------------
> Do You Yahoo!? -- Une adresse @yahoo.fr gratuite et en français !
> Testez le nouveau Yahoo! Mail