[Insight-users] LabelMaps Handling and Colouring

[Adrian Friebel Work]

Hi,
I have some questions, regarding the LabelMaps/Objects and their 
functionalities. I am not sure, if the solutions, that I see for my 
problem, are okay, or if their are better ways to do what I want.

Here is my problem:
After some preprocessing on the (separated) three color channels of an 
RGB image, I used the BinaryImageToShapeLabelMapFilter on the red 
channel to get a LabelMap of the objects of interest. I used shape 
opening to remove the biggest and smallest objects. To this point 
everything looks nice.
Now, I want to distinguish these LabelObjects into two classes, based on 
the binarized values in the (at this stage separated) blue channel.
After this, I want to generate an RGB overlay image of the original 
image (grey scale) and the two different LabelObject classes. At best, 
with two different colours for the two different LabelObject classes. (I 
used the LabelOverlayImageFilter to overlay the original image with all 
LabelObjects, which worked fine)


So, I thought about, to convert the ShapeLabelMap to an 
AttributeLabelMap. Then I could look for example at their centroid point 
in the blue channel image and based on this, I set my LabelObject 
attribute value to true or false. (Similar to the 
BinaryReconstructionLabelMapFilter, I guess) But is it possible, to use 
the Attribute values, to generate the colour for the overlay?

The second way to go, that I see, is, to create a second LabelMap and 
copy the LabelObjects with blue at their centroid point to this new 
LabelMap and delete them from the old. (I tried this, but my application 
crashed during usage of functions like LabelObject->CopyAllFrom(obj) and 
LabelMap->RemoveLabelObject(obj)). Is this possible in principle, how 
can I copy and delete LabelObjects to/from LabelMaps? And if I have the 
two LabelMaps, is it possible to lay them both over the input image and 
give the LabelObjects of the two different LabelMaps two different colours?

I would be grateful for some hints and approach sketches,
thanks in advance,
Adrian.