itkGaussianMixtureModelComponent.txx

Go to the documentation of this file.

/*=========================================================================

Program:   Insight Segmentation & Registration Toolkit
Module:    $RCSfile: itkGaussianMixtureModelComponent.txx,v $
Language:  C++
Date:      $Date: 2009-04-16 15:27:01 $
Version:   $Revision: 1.19 $

Copyright (c) Insight Software Consortium. All rights reserved.
See ITKCopyright.txt or http://www.itk.org/HTML/Copyright.htm for details.

This software is distributed WITHOUT ANY WARRANTY; without even 
the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR 
PURPOSE.  See the above copyright notices for more information.

=========================================================================*/

#ifndef __itkGaussianMixtureModelComponent_txx
#define __itkGaussianMixtureModelComponent_txx

#include <iostream>

#include "itkGaussianMixtureModelComponent.h"

namespace itk { 
namespace Statistics {
  
template< class TSample >
GaussianMixtureModelComponent< TSample >
::GaussianMixtureModelComponent()
{
  m_MeanEstimator = MeanEstimatorType::New();
  m_CovarianceEstimator = CovarianceEstimatorType::New();
  m_GaussianDensityFunction = NativeMembershipFunctionType::New();
  this->SetMembershipFunction((MembershipFunctionType*)
                              m_GaussianDensityFunction.GetPointer());
  m_Mean.Fill(0.0);
  m_Covariance.SetIdentity();
}

template< class TSample >
void
GaussianMixtureModelComponent< TSample >
::PrintSelf(std::ostream& os, Indent indent) const
{
  Superclass::PrintSelf(os, indent);

  os << indent << "Mean: " << m_Mean << std::endl;
  os << indent << "Covariance: " << m_Covariance << std::endl;
  os << indent << "Mean Estimator: " << m_MeanEstimator << std::endl;
  os << indent << "Covariance Estimator: " << m_CovarianceEstimator << std::endl;
  os << indent << "GaussianDensityFunction: " << m_GaussianDensityFunction << std::endl;
}

template< class TSample >
void
GaussianMixtureModelComponent< TSample >
::SetSample(const TSample* sample)
{
  Superclass::SetSample(sample);

  m_MeanEstimator->SetInputSample(sample);
  m_CovarianceEstimator->SetInputSample(sample);

  WeightArrayType* weights = this->GetWeights();
  m_MeanEstimator->SetWeights(weights);
  m_CovarianceEstimator->SetWeights(weights);
  const MeasurementVectorSizeType measurementVectorLength = 
            sample->GetMeasurementVectorSize();
  m_GaussianDensityFunction->SetMeasurementVectorSize( 
                                    measurementVectorLength );
  MeasurementVectorTraits::SetLength( m_Mean, measurementVectorLength );
  m_Covariance.SetSize( measurementVectorLength, measurementVectorLength );
  m_Mean.Fill(NumericTraits< double >::NonpositiveMin());
  m_Covariance.Fill(NumericTraits< double >::NonpositiveMin());
  m_CovarianceEstimator->SetMean(&m_Mean);
  m_GaussianDensityFunction->SetMean(&m_Mean);
}

template< class TSample >
void
GaussianMixtureModelComponent< TSample >
::SetParameters(const ParametersType &parameters)
{
  Superclass::SetParameters(parameters);

  unsigned int paramIndex = 0;
  unsigned int i, j;

  bool changed = false;

  MeasurementVectorSizeType measurementVectorSize = 
          this->GetSample()->GetMeasurementVectorSize();

  for ( i = 0; i < measurementVectorSize; i++)
    {
    if ( m_Mean[i] != parameters[paramIndex] )
      {
      m_Mean[i] = parameters[paramIndex];
      changed = true;
      }
    ++paramIndex;
    }

  for ( i = 0; i < measurementVectorSize; i++ )
    {
    for ( j = 0; j < measurementVectorSize; j++ )
      {
      if ( m_Covariance.GetVnlMatrix().get(i, j) != 
           parameters[paramIndex] )
        {
        m_Covariance.GetVnlMatrix().put(i, j, parameters[paramIndex]);
        changed = true;
        }
      ++paramIndex;
      }
    }
  m_GaussianDensityFunction->SetCovariance(&m_Covariance);
  this->AreParametersModified(changed);
}


template< class TSample >
double
GaussianMixtureModelComponent< TSample >
::CalculateParametersChange()
{
  unsigned int i, j;

  MeanType meanEstimate = *(m_MeanEstimator->GetOutput());
  CovarianceType covEstimate = *(m_CovarianceEstimator->GetOutput());

  double temp;
  double changes = 0.0;
  MeasurementVectorSizeType measurementVectorSize = 
          this->GetSample()->GetMeasurementVectorSize();
  
  for ( i = 0; i < measurementVectorSize; i++)
    {
    temp = m_Mean[i] - meanEstimate[i];
    changes += temp * temp;
    }

  for ( i = 0; i < measurementVectorSize; i++ )
    {
    for ( j = 0; j < measurementVectorSize; j++ )
      {
      temp = m_Covariance.GetVnlMatrix().get(i, j) - 
        covEstimate.GetVnlMatrix().get(i, j);
      changes += temp * temp;
      }
    }

  changes = vcl_sqrt(changes);
  return changes;
}

template< class TSample >
void
GaussianMixtureModelComponent< TSample >
::GenerateData()
{
  MeasurementVectorSizeType measurementVectorSize = 
          this->GetSample()->GetMeasurementVectorSize();
  
  this->AreParametersModified(false);

  m_MeanEstimator->Update();

  unsigned int i, j;
  double temp;
  double changes;
  bool changed = false;
  ParametersType parameters = this->GetFullParameters();
  int paramIndex  = 0;

  MeanType meanEstimate = *(m_MeanEstimator->GetOutput());
  for ( i = 0; i < measurementVectorSize; i++)
    {
    temp = m_Mean[i] - meanEstimate[i];
    changes = temp * temp;
    changes = vcl_sqrt(changes);
    if ( changes > this->GetMinimalParametersChange() )
      {
      changed = true;
      }
    }


  if ( changed )
    {
    m_Mean = *(m_MeanEstimator->GetOutput());
    for ( i = 0; i < measurementVectorSize; i++)
      {
      parameters[paramIndex] = meanEstimate[i];
      ++paramIndex;
      }
    this->AreParametersModified(true);
    }
  else
    {
    paramIndex = measurementVectorSize;
    }

  m_CovarianceEstimator->Update();
  CovarianceType covEstimate = *(m_CovarianceEstimator->GetOutput());
  changed = false;
  for ( i = 0; i < measurementVectorSize; i++ )
    {
    for ( j = 0; j < measurementVectorSize; j++ )
      {
      temp = m_Covariance.GetVnlMatrix().get(i, j) - 
        covEstimate.GetVnlMatrix().get(i, j);
      changes = temp * temp;
      changes = vcl_sqrt(changes);
      if ( changes > this->GetMinimalParametersChange() )
        {
        changed = true;
        }
      }
    }
  
  if ( changed )
    {
    m_Covariance = *(m_CovarianceEstimator->GetOutput());
    for ( i = 0; i < measurementVectorSize; i++ )
      {
      for ( j = 0; j < measurementVectorSize; j++ )
        {
        parameters[paramIndex] = covEstimate.GetVnlMatrix().get(i, j);
        ++paramIndex;
        }
      }
    this->AreParametersModified(true);
    }

  Superclass::SetParameters(parameters);
  //update covariance and its inverse of Gaussian mixture
  m_GaussianDensityFunction->SetCovariance( &m_Covariance ); 
}
    
} // end of namespace Statistics 
} // end of namespace itk 

#endif