[Insight-users] Question on ImageKmeansModelEstimator

[Uwe Köhler]

Hi folks,

while digging deeper into the ImageKmeansModelEstimator code I was wondering 
why the class is templated over TMembershipFunction. The only place that 
seems to be used is right after everything has finished and the found 
centroids are assigned to the membership functions. I was left to believe 
initially, that the memebership function is used during the Kmeans algorithm 
(this is what I am planning to do). At the moment the distance measure is 
implemented in NearestNeighborSearchBasic as an Euklidien distance. I will 
try to change it to the TMembershipFunction. Would anybody be interested? 
Should it be made clearer in the documentation?

Any thoughts are very wellcome

Uwe