[Insight-users] Image statistics on regions of an image

[Miller, James V (Research)]

Andre, 

You should take a look at LabelStatisticsImageFilter.  Given an image
and a label map, it calculates statistics for pixels under each label.

After running the filter, you can query the 

	Minimum
	Maximum
	Mean
	Sigma
	Variance
	Sum
	Count

for a particular label.

If you need more "features" than these, please let us know.  We could 
extend the filter or create a subclass that calculates other features
over the regions.

Jim


-----Original Message-----
From: Luis Ibanez [mailto:luis.ibanez at kitware.com]
Sent: Sunday, November 07, 2004 11:27 AM
To: ahuisman at cs.uu.nl; 'insight-users at itk.org'
Subject: Re: [Insight-users] Image statistics on regions of an image



Hi Andre

Your request was logged into the buck tracking database

  Bug 1331 (Feature Request).
  http://www.itk.org/Bug/bug.php?op=show&bugid=1331&pos=0

and as a first response we added a class

      itkMinimumMaximumConditionalImageCalculator

in

                Insight/Code/BasicFilters

This class computes the Minimum and Maximum values
of an image but considering only the pixels for which
a second image has a particular pixel value.

In practice you provide two images to this class. The first
input is the image from which you want to compute the
statistics, while the second input is the class that defines
what pixels to use.

You can use your labeled mask (the result of your segmentation)
as second input to this class.

Please update your CVS checkout in order to get this new class.


You probably will need other types of statistics for your images.
If so, the generic way to go will be to add a Conditional Functor
to the classes in Code/Numerics/Statistics.   You could then create
your own conditional Functors and instantiate any of the classes
responsible for computing Histograms.

A Conditional Functor will be a class that has a Test() method
returning a Boolean. We could make Functor to be a template
parameter of the statistical classes, just the way the
UnaryFunctorImageFilter is implemented.

Please let us know about types of statistics that you would like
to compute from your cell images, and whether these images
are grayscale, Color, or multi-band.


    Thanks


         Luis


===========================================
*Andre Huisman* wrote:

For our research on texture features describing changes in the
chromatin distribution in cell nuclei we need to do calculations on
cells. We already have segmentation results (stored in a grey-scale
image where each cell has a unique constant grey-value representing its
number), but now we want to compute statistics on those separate cells.
However, I can not find a method in ITK to compute several statistics
(for example using <a class="el"
 
href="http://www.itk.org/Doxygen/html/classitk_1_1MinimumMaximumImageFilter.
html">itk::MinimumMaximumImageFilter</a>
or <a class="el"
 
href="http://www.itk.org/Doxygen/html/classitk_1_1MinimumMaximumImageCalcula
tor.html">itk::MinimumMaximumImageCalculator</a>
and <a class="el"
 
href="http://www.itk.org/Doxygen/html/classitk_1_1StatisticsImageFilter.html
">itk::StatisticsImageFilter</a>)
using a masked image. Of course I can make a bounding box around each
cell and than compute those statistics, but that will introduce a large
error for most statistics. <br>
Are there any possibilities in ITK to use a mask to restrict
computations on images to certain regions?<br>
<br>
Thank you in advance,<br>
<br>
Andr&eacute; Huisman<br>
University Medical Center Utrecht, the Netherlands<br>
Department of pathology<br>





_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users