Package gov.nih.mipav.model.algorithms.registration

Class RegistrationMeasureNormalizedMutualInformation

  • public class RegistrationMeasureNormalizedMutualInformation
extends RegistrationMeasure
    Concrete implementation of the RegistrationMeasure class based on the normalized mutual information computed as follows:

    H(X,Y) / (H(X) + H(Y))

    where

    - X and Y are the target and registered source images, respectively, or vice versa - H(X) and H(Y) are the standard marginal entropy definition where H(X) = - sum_i { p_i * log(p_i) } for probability p_i in histogram bin i - H(X,Y) is the joint entropy definition where H(X,Y) = - sum_ij { p_ij * log(p_ij) } for joint probability p_ij in histogram bin (i,j).

    • Field Summary

      Fields 
      Modifier and Type Field Description
      protected int[] m_aiBinReg
      Bins used to hold the histograms.
      protected int[] m_aiBinRegTrg
      DOCUMENT ME!
      protected int[] m_aiImageBinReg
      Image corresponding to the samples in the registered source image which identifies to which bin the sample belongs.
      protected int[] m_aiImageBinTrg
      Image corresponding to the samples in the target image which identifies to which bin the sample belongs.
      protected double m_dEntropyImageTrg
      Precomputed entropy for the target image.
      protected float m_fSrcMin
      Given a value in the registered source image, these values are used in this equation to determine the actual histogram bin: int iBinReg = (int)((value - m_fSrcMin) * m_fSrcScale);.
      protected float m_fSrcScale
      DOCUMENT ME!
      protected int m_iNumBins
      Number of bins to use for computation.
      protected static double[] s_adIntLog
      Table with precalculated logarithms.

    • Method Summary

      All Methods Static Methods Instance Methods Concrete Methods 
      Modifier and Type Method Description
      RegistrationMeasure createNew()
      Create a new instance of this measure.
      protected static double getEntropy​(int[] aiHistBin, double dNumSamples)
      Standard entropy calculation.
      double getError()
      Return the current error measure between the target image and the registered source image.
      java.lang.String getName()
      Return the name of the particular registration measure.
      static java.lang.String getStaticName()
      Return the name of the particular registration measure without having an instance.
      protected static double[] getTableIntLog()
      DOCUMENT ME!
      void setImages​(ModelSimpleImage kImageSrc, ModelSimpleImage kImageTrg)
      Setup the measure to use information from the target image and the original source image.
      void updateRegistration​(int iIndex, float fNewValue)
      Update the specified value in the registered source image.

      • Methods inherited from class java.lang.Object

        clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

    • Field Detail

      • s_adIntLog

        protected static double[] s_adIntLog
        Table with precalculated logarithms.

      • m_aiBinReg

        protected int[] m_aiBinReg
        Bins used to hold the histograms.

      • m_aiBinRegTrg

        protected int[] m_aiBinRegTrg
        DOCUMENT ME!

      • m_aiImageBinReg

        protected int[] m_aiImageBinReg
        Image corresponding to the samples in the registered source image which identifies to which bin the sample belongs. This is a linear storage for a 2D array where iBinReg and iBinTrg map to the 1D index by computing iBinReg*m_iNumBins+iBinTrg.

      • m_aiImageBinTrg

        protected int[] m_aiImageBinTrg
        Image corresponding to the samples in the target image which identifies to which bin the sample belongs.

      • m_dEntropyImageTrg

        protected double m_dEntropyImageTrg
        Precomputed entropy for the target image.

      • m_fSrcMin

        protected float m_fSrcMin
        Given a value in the registered source image, these values are used in this equation to determine the actual histogram bin: int iBinReg = (int)((value - m_fSrcMin) * m_fSrcScale);.

      • m_fSrcScale

        protected float m_fSrcScale
        DOCUMENT ME!

      • m_iNumBins

        protected int m_iNumBins
        Number of bins to use for computation.

    • Constructor Detail

      • RegistrationMeasureNormalizedMutualInformation

        public RegistrationMeasureNormalizedMutualInformation()
        Constructor which sets up for a default number of bins.

      • RegistrationMeasureNormalizedMutualInformation

        public RegistrationMeasureNormalizedMutualInformation​(int iNumBins)
        Constructor.
        Parameters:
        iNumBins - int Number of bins to use for computation.

    • Method Detail

      • getStaticName

        public static java.lang.String getStaticName()
        Return the name of the particular registration measure without having an instance.
        Returns:
        String Contains the name of the particular registration measure.

      • createNew

        public RegistrationMeasure createNew()
        Create a new instance of this measure. That is, the same type and having the same properties. But initially, the target is not defined.
        Specified by:
        createNew in class RegistrationMeasure
        Returns:
        RegistrationMeasure Newly created instance.

      • getError

        public double getError()
        Return the current error measure between the target image and the registered source image.
        Specified by:
        getError in class RegistrationMeasure
        Returns:
        double Error measure value. Minimum value is zero. Maximum value depends on the range of values in the target and registered source images.

      • getName

        public java.lang.String getName()
        Return the name of the particular registration measure.
        Specified by:
        getName in class RegistrationMeasure
        Returns:
        String Contains the name of the particular registration measure.

      • setImages

        public void setImages​(ModelSimpleImage kImageSrc,
                      ModelSimpleImage kImageTrg)
        Setup the measure to use information from the target image and the original source image. Each derived class override of this method must call this method.
        Overrides:
        setImages in class RegistrationMeasure
        Parameters:
        kImageSrc - ModelSimpleImage Contains information and data about the original source image. This array may be empty but not null.
        kImageTrg - ModelSimpleImage Contains information and data about yhe target image. This array may be empty but not null.

      • updateRegistration

        public void updateRegistration​(int iIndex,
                               float fNewValue)
        Update the specified value in the registered source image. Causes a change in the computed error measure.
        Specified by:
        updateRegistration in class RegistrationMeasure
        Parameters:
        iIndex - int Linear array index for the value to be updated.
        fNewValue - float New registered source image value to store at the specified index.

      • getEntropy

        protected static double getEntropy​(int[] aiHistBin,
                                   double dNumSamples)
        Standard entropy calculation.
        Parameters:
        aiHistBin - int[] Array of histogram bins. The sum of the the values in each of the array positions should equal the values passed in iNumSamples.
        dNumSamples - double Total number of samples in the distribution. Must be a double-precision floating point number because the number can be quite large in the case of the joint entropy.
        Returns:
        double Computed entropy value.

      • getTableIntLog

        protected static double[] getTableIntLog()
        DOCUMENT ME!
        Returns:
        DOCUMENT ME!