[Insight-users] connected components as seeds for fastmarching

[Tim Rudge]

Hi again,

I'd like to have a filter that given a binary image will produce a node for 
each connected group of non-zero voxels (so that I can pass them as seeds 
to fastmarchingimagefilter to produce a distance map for the 
ShapeDetectionLevelSetImageFilter).

I have tried simply passing each non-zero location as a seed with initial 
distance of zero - this gives wierd unexpected results, perhaps because 
some of them are neighbouring each other? I figure if I only pass a single 
seed node for each connected group this will work? Or perhaps only non-zero 
points which have a zero neighbour and are therefore on the border of a 
non-zero region?

If this is a reasonable approach, how can I implement this, pref. without 
writing my own filter from scratch.

The background to this is that we have confocal microscope images where the 
regions we want are dark, seperated by bright boundaries. I want to use a 
threshold to find points in the dark regions and start the level sets 
around these points.

Many thanks,

Tim

Tim Rudge
---------
Cellular Development Lab
Dept. Plant Sciences
Cambridge University