[Insight-users] RGB Segmentation

Luis Ibanez luis.ibanez at kitware.com
Mon Aug 2 23:17:51 EDT 2004


Hi Jacob,


1) Yes, you can do segmentation of color images in ITK
         Please look at the VectorConfidenceConnectedImageFilter
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VectorConfidenceConnectedImageFilter.html

         and its example under

         /src/Insight/Examples/Segmentation/
                          VectorConfidenceConnected.cxx

         Most of the anatomical structures shown in the cover
         of the ITK SofwareGuide were segmented with this filter
         from the RGB VisibleWoman dataset.

         In this filter you specify seed points, and their
         vector-pixel values are used for defining a Mahalanobis
         distance in the RGB space.



2) Yes, you can compose a color image between the
         original image and the segmented region.

         For composing the image you want to write an
         ImageAdaptor and define your strategy for
         blending the mask and the color image.

         You will find ImageAdaptors described in detail
         in the ITK Software Guide.

           http://www.itk.org/ItkSoftwareGuide.pdf


         You could also implement a blending filter
         by using the BinaryFunctorFilter.

http://www.itk.org/Insight/Doxygen/html/classitk_1_1BinaryFunctorImageFilter.html



Regards,


     Luis



--------------------------
Jacob Boomgaarden wrote:

> Hello Everyone,
> 
> I am working with the ConfidenceConnectedImageFilter to do segmentation of both
> lungs in a two-dimensional CT image. The image I am currently working with has pixel
> values at 8-bit grayscale and is a .png image file.
> 
> I wish to now do the same type of segmentation on a 2D colored cryosection that has
> pixel values that are 24-bit RGB colored and is also in a .png image file. My
> question is if its possible to the segmentation on this colored image and keep an
> RGB color to the segmented region. Such things that deal with this that I am curious
> on are: Should I define my image type using RGBPixel? When using the
> SetReplaceValue() method of the ConfidenceConnectedImageFilter class, how do I
> specify that the pixeltype I want to send in has a certain intensity for each RGB
> value...can this be done simply by setting up an RGBPixel instance, giving it
> specific RGB values and then passing it in as a reference?
> 
> Thanks for your help,
> 
> 
> 






More information about the Insight-users mailing list