[Insight-users] RegionGrowingSegmentation

[Luis Ibanez]

Hi Suresh,

The RegionGrowingSegmentaion example process the input
image in two steps. First it applies an edge-preserving
smoothing. You could see the result of this smoothing
by clicking on the button "Homogeneous Image".

The parameters of the "AnisotropicDiffusion" methods
are nicely described in:
http://www.itk.org/Insight/Doxygen/html/classitk_1_1AnisotropicDiffusionFunction.html
take a look at the section "Overview of anisotropic diffusion".

The number of iteration will increase the smoothing
and also increase linearly the processing time.
The recomended timestep should probably be fixed to 0.25
for 2D images and 0.125 for 3D images. Larger timesteps
will result in increased smoothing but could make
the computation unstable.  The conductance term will
control the degree of smoothing. A larger conductance
will result in increased smoothing.


The second step applies a region growing technique.
This requires you to provide a seed point by clicking
on the image with the mouse left button. The coordinate
of the selected pixel should appear on the GUI as soon
as you click. They are displayed on the text box below
the "Homogeneous Image" button. You should click on the
image labeled "homogeneous image", not on the input image.

The region growing algorithms will extract a set of
pixels connected to the seed following a particular
condition.  In the case of the ThresholdConnected
the criterion is that the pixels on the set should
have gray levels in a range specified by the user.
(you can set this range from the GUI using the
text boxes below the "ConnectedThreshold" button.
In the case of the "ConfidenceConnected" algorithm,
the criterion is that the gray level of the pixel
should belong to a range defined by the variance
of the current region. This last is repeated on
several iterations (specified by the user too).

Make sure that you selected the seed point before
you trigger any of the region growing algorithms.

Please let us know if you encounter any problem,

Thanks

Luis


==========================


suresh wrote:
> Hi Luis,
> Hi Sayan,
> 
> Thanks for your help. I'm working the Segmentation problem using the 
> RegionGrowingSegmentation example sugggested by Luis.
> But unfortunately what i'm getting as result is just a blank image.
> My input is a MRI of 122 slices.Whats could be the problem with my 
> code.?? pleae help me with this.
> Also i would like to know about the following parameters of 
> GradientAnisotropicDiffusionImageFilterType.
> 
> 1. ConductanceParameter
> 2. TimeStep
> 
> what is their significance? How does they effect the Segmentation process?
> please help me.
> 
> thanks
> 
> suresh
>