[Insight-users] ExpectationMaximizationMixtureModelEstimator in 3D

[smrolfe at u.washington.edu]

Hello, 

I've been debugging my implementation of the ExpectationMaximizationMixtureModelEstimator.  I just found that when I adapt the example ExpectationMaximizationMixtureModelEstimator_2D.cxx to 3D that I generate the same error: 

File: c:\users\smrolfe\itk\insighttoolkit-4.3.1\modules\numerics\statistics\include\itkWeightedMeanSampleFilter.hxx
Line: 149
Description: itk::ERROR: WeightedMeanSampleFilter(0000000000B24F00): Total weight was too close to zero. Value = 0

Is there a problem implementing this filter in 3D?  Are there any examples of this? I am assuming that the parameter assignment would be:

  params[0] // mean of dimension 1
  params[1] // mean of dimension 2
  params[2] // mean of dimension 3
  params[3] // covariance(0,0)
  params[4] // covariance(0,1)
  params[5] // covariance(0,2)
  params[6] // covariance(1,0)
  params[7] // covariance(1,1)
  params[8] // covariance(1,2)
  params[9] // covariance(2,0)
  params[10]// covariance(2,1)
  params[11]// covariance(2,2) 

I appreciate your help.
Thanks,
Sara