Class for floating point valued histograms E.g. needed to represent an normalized histogram when performing histogram equalization WARNING: currently only works for default case when bin size is 1. More...

#include <FloatHistogram.h>

Inheritance diagram for stira::histogram::FloatHistogram:

Public Member Functions
	FloatHistogram (int nrOfBins, int nrBands)
	constructor for histogram object More...

	FloatHistogram (IntHistogram *pIntHistogram, bool requiresNormalization)
	constructor with initialization from IntHistogram More...

	FloatHistogram (const FloatHistogram &pFloatHistogram)
	Copy constructor. More...

FloatHistogram *	Clone ()
	Creates a clone of this FloatHistogram This clone becomes the responsability of the caller to delete.

bool	UpdateHistogram (IntHistogram *pIntHistogram, double alpha)

virtual	~FloatHistogram ()
	destructor for histogram object

void	Normalize ()
	Normalizes each band in the current histogram to 1.

double	GetBinValueHistogramData (int band, int bin)
	Gets the value that corresponds with the given band and bin. More...

void	InitializeWithNormalizedHistogram (IntHistogram *pStandardHistogram)
	compute normalized histogram by dividing the histogram values by the total number of counts in the histogram More...

void	InitializeWithNormalizedCumulativeHistogram (IntHistogram pStandardHistogram, IntHistogram pCumulativeHistogram)
	compute normalized cumulative histogram by dividing the cumulative histogram values by the standard histogram values More...

double	ComputeStandardKullbackLeiblerDistance (FloatHistogram *pOtherHistogram)
	computes standard Kullback-Leibler divergence between this histogram and the histogram in the argument Reference http://en.wikipedia.org/wiki/Kullback–Leibler_divergence Zero entries: http://mathoverflow.net/questions/72668/how-to-compute-kl-divergence-when-pmf-contains-0s Warning: assumes that both histograms are normalized More...

double	ComputeEntropy ()
	computes entropy of this histogram Reference: http://en.wikipedia.org/wiki/Entropy_(information_theory) Zero entries: http://mathoverflow.net/questions/72668/how-to-compute-kl-divergence-when-pmf-contains-0s Warning: assumes that this histogram is normalized

double	ComputeSymmetricKullbackLeiblerDistance (FloatHistogram *pOtherHistogram)
	computes symmetric Kullback-Leibler divergence between this histogram and the histogram in the argument Reference http://en.wikipedia.org/wiki/Kullback–Leibler_divergence Warning: assumes that both histograms are normalized More...

double	ComputeJensenShannonMetric (FloatHistogram *pOtherHistogram)
	computes the square root of the Jensen–Shannon divergence between the current histogram and the histogram passed as argument the Jensen–Shannon divergence is a popular method of measuring the similarity between two probability distributions. It is based on the Kullback–Leibler divergence, with the notable (and useful) difference that it is always a finite value. The square root of the Jensen–Shannon divergence is a metric. Reference: - http://en.wikipedia.org/wiki/Jensen–Shannon_divergence More...

double	ComputeIDivergence (FloatHistogram *pOtherHistogram)
	computes Csiszar's I-divergence between this histogram and the histogram in the argument Reference Csiszar, I. (1991). "Why least squares and maximum entropy? an axiomatic approach to inference for linear inverse problems. Annals of Statistics, 19(4):2032–2066. See online access at http://www.jstor.org/ More...

virtual bool	Write (std::string fileName)
	writes values from current histogram to a file More...

void	SetEpsilonThreshold (double epsilonValue)
	Sets the threshold value to compare values in the histogram with each other or with zero. More...

double	GetEpsilonThreshold ()
	Gets the threshold value to compare values in the histogram with each other or with zero.

Public Member Functions inherited from stira::histogram::HistogramMaster< double >
void	AddDataPoint (int valueToAdd, int bandNr=0)
	adds a value to the histogram This is a generalization of AddOneToBin, since here also bin sizes not equal to one can be used; the bin number will be calculated depending on the bin size and then one will be added to that bin. More...

void	SetBinsToZero ()
	set all bins to zero

double	GetBinValue (int bandNr, int binNr) const
	gets nr of counts in given bin nr for a given band nr if the macro USE_SAFE is defined, we use the version with boundary checking, else we use the fast (unsafe) call to get the data from the array More...

double	GetBinValueSafe (int bandNr, int binNr) const
	gets nr of counts in given bin nr for a given band nr Specific safe call for above with extra boundary checking More...

int	GetNrOfBins () const
	Gets nr of bins in the histogram For now, we assume bin size is 1, so the number of bins is the data range mRange.

int	GetNrOfBands () const
	Gets nr of bands in the histogram.

int	GetNrOfCounts () const
	Gets total nr of counts in the histogram This number is the number of pixels in the image that were counted. We don't count the same pixel over and over again for different bands in the image; we just assume that the ROI for which the histogram was computed is the same for all bands in an image.

double	GetBinSize () const
	Gets the width of a histogram bin.

double	GetDataMinimum () const
	Gets the minimum of the data range.

double	GetDataMaximum () const
	Gets the maximum of the data range.

void	SetBinValue (int bandNr, int binNr, doublevalue)
	sets nr of counts in given bin nr for a given band nr if the macro USE_SAFE is defined, we use the version with boundary checking, else we use the fast (unsafe) call to get the data from the array More...

Additional Inherited Members
Protected Member Functions inherited from stira::histogram::HistogramMaster< double >
void	AddOneToBin (int bandNr, int binNr)
	adds one count in given bin nr for a given band nr if the macro USE_SAFE is defined, we use the version with boundary checking, else we use the fast (unsafe) call to get the data from the array More...

void	AddOneToBin (int binNr)
	adds one count in given bin nr in case there is just a single band if the macro USE_SAFE is defined, we use the version with boundary checking, else we use the fast (unsafe) call to get the data from the array More...

void	AddOneToBinSafe (int bandNr, int binNr)
	adds one count in given bin nr for a given band nr Specific safe call for above with extra boundary checking More...

void	AddOneToBinSafe (int binNr)
	adds one count in given bin nr in case there is just a single band Specific safe call for above with extra boundary checking More...

	HistogramMaster ()
	constructor only to be called from child classes

	HistogramMaster (const HistogramMaster &pHistogram)
	Copy constructor. More...

virtual	~HistogramMaster ()
	destructor only to be called from child classes

void	Initialize (doublebinSize, int nrOfBands, doubleminValue, doublemaxValue)
	Initializes member variables and creates and initializes histogram data structure Creation and initialization of the histogram data structure is done by calling CreateInitializedDataStructure, which does not initialize the other member variables. More...

void	CreateInitializedDataStructure ()
	Just creates and initializes histogram data structure does not initialize the other member variables.

void	SetBinValueSafe (int bandNr, int binNr, doublevalue)
	sets nr of counts in given bin nr for a given band nr Specific safe call for above with extra boundary checking More...

Protected Attributes inherited from stira::histogram::HistogramMaster< double >
double **	mpHistogram
	the actual histogram (we count nr of intensities per colour band)

double	mBinSize
	size of histogram bin

int	mNrOfBands
	nr of bands in image

int	mNrOfBins
	nr of bins in histogram

int	mNrOfCounts
	nr of pixels in image = nr of counts in histogram per band

double	mMin
	min intensity in histogram

double	mMax
	max intensity in histogram

Detailed Description

Class for floating point valued histograms E.g. needed to represent an normalized histogram when performing histogram equalization WARNING: currently only works for default case when bin size is 1.

Constructor & Destructor Documentation

stira::histogram::FloatHistogram::FloatHistogram	(	int	nrOfBins,
		int	nrBands
	)

constructor for histogram object

Parameters

nrOfBins	number of bins in the histogram
nrBands	number of bands in the histogram (on per colour chanel)

stira::histogram::FloatHistogram::FloatHistogram	(	IntHistogram *	pIntHistogram,
		bool	requiresNormalization
	)

constructor with initialization from IntHistogram

Parameters

pIntHistogram	histogram to initialize this histogram from
requiresNormalization	flag whether the values need to be normalized

References stira::histogram::HistogramMaster< T >::GetBinValue(), stira::histogram::HistogramMaster< T >::GetNrOfBands(), and stira::histogram::HistogramMaster< T >::GetNrOfBins().

stira::histogram::FloatHistogram::FloatHistogram ( const FloatHistogram & pFloatHistogram )

inline

Copy constructor.

Parameters

pFloatHistogram histogram from which to initialize this new histogram

References Clone(), ComputeEntropy(), ComputeIDivergence(), ComputeJensenShannonMetric(), ComputeStandardKullbackLeiblerDistance(), ComputeSymmetricKullbackLeiblerDistance(), GetBinValueHistogramData(), GetEpsilonThreshold(), InitializeWithNormalizedCumulativeHistogram(), InitializeWithNormalizedHistogram(), Normalize(), SetEpsilonThreshold(), Write(), and ~FloatHistogram().

Member Function Documentation

double stira::histogram::FloatHistogram::ComputeIDivergence ( FloatHistogram * pOtherHistogram )

computes Csiszar's I-divergence between this histogram and the histogram in the argument Reference Csiszar, I. (1991). "Why least squares and maximum entropy? an axiomatic approach to inference for linear inverse problems. Annals of Statistics, 19(4):2032–2066. See online access at http://www.jstor.org/

Parameters

pOtherHistogram histogram to compute distance with

References stira::histogram::HistogramMaster< T >::GetBinValue(), stira::histogram::HistogramMaster< T >::GetNrOfBands(), and stira::histogram::HistogramMaster< T >::GetNrOfBins().

Referenced by FloatHistogram().

double stira::histogram::FloatHistogram::ComputeJensenShannonMetric ( FloatHistogram * pOtherHistogram )

computes the square root of the Jensen–Shannon divergence between the current histogram and the histogram passed as argument the Jensen–Shannon divergence is a popular method of measuring the similarity between two probability distributions. It is based on the Kullback–Leibler divergence, with the notable (and useful) difference that it is always a finite value. The square root of the Jensen–Shannon divergence is a metric. Reference: - http://en.wikipedia.org/wiki/Jensen–Shannon_divergence

A New Metric for Probability Distributions, D. Endres & J. Schindelin, IEEE Trans. Information Theory, VOL. 49, NO. 7, JULY 2003, pp. 1858 – 1860
Parameters

pOtherHistogram histogram to compute distance with

References ComputeStandardKullbackLeiblerDistance(), stira::histogram::HistogramMaster< T >::GetBinValue(), and stira::histogram::HistogramMaster< T >::SetBinValue().

Referenced by stira::filter::LocalBinaryPattern::ComputeJensenShannonDivergenceBetweenImagePatches(), and FloatHistogram().

double stira::histogram::FloatHistogram::ComputeStandardKullbackLeiblerDistance ( FloatHistogram * pOtherHistogram )

computes standard Kullback-Leibler divergence between this histogram and the histogram in the argument Reference http://en.wikipedia.org/wiki/Kullback–Leibler_divergence Zero entries: http://mathoverflow.net/questions/72668/how-to-compute-kl-divergence-when-pmf-contains-0s Warning: assumes that both histograms are normalized

Parameters

pOtherHistogram histogram to compute distance with

References stira::histogram::HistogramMaster< T >::GetBinValue(), stira::histogram::HistogramMaster< T >::GetNrOfBands(), and stira::histogram::HistogramMaster< T >::GetNrOfBins().

Referenced by ComputeJensenShannonMetric(), and FloatHistogram().

double stira::histogram::FloatHistogram::ComputeSymmetricKullbackLeiblerDistance ( FloatHistogram * pOtherHistogram )

computes symmetric Kullback-Leibler divergence between this histogram and the histogram in the argument Reference http://en.wikipedia.org/wiki/Kullback–Leibler_divergence Warning: assumes that both histograms are normalized

Parameters

pOtherHistogram histogram to compute distance with

References stira::histogram::HistogramMaster< T >::GetBinValue(), stira::histogram::HistogramMaster< T >::GetNrOfBands(), and stira::histogram::HistogramMaster< T >::GetNrOfBins().

Referenced by FloatHistogram().

double stira::histogram::FloatHistogram::GetBinValueHistogramData	(	int	band,
		int	bin
	)

Gets the value that corresponds with the given band and bin.

Parameters

band	the image band for which to get the histogram value
bin	the bin for which to get the histogram value

Returns: value that corresponds with band and bin

Referenced by FloatHistogram().

void stira::histogram::FloatHistogram::InitializeWithNormalizedCumulativeHistogram	(	IntHistogram *	pStandardHistogram,
		IntHistogram *	pCumulativeHistogram
	)

compute normalized cumulative histogram by dividing the cumulative histogram values by the standard histogram values

Parameters

pStandardHistogram	standard histogram, needed to normalize the cumulative histogram
pCumulativeHistogram	histogram to be normalized

References stira::histogram::HistogramMaster< T >::GetBinValue(), and stira::histogram::HistogramMaster< T >::GetNrOfCounts().

Referenced by FloatHistogram(), stira::contrastenhance::HistogramTools::HistogramEqualizeSingleBand(), and stira::contrastenhance::AdaptiveHistogramEqualizer::Run().

void stira::histogram::FloatHistogram::InitializeWithNormalizedHistogram ( IntHistogram * pStandardHistogram )

compute normalized histogram by dividing the histogram values by the total number of counts in the histogram

Parameters

pStandardHistogram standard histogram, needed to normalize the cumulative histogram

References stira::histogram::HistogramMaster< T >::GetBinValue(), and stira::histogram::HistogramMaster< T >::GetNrOfCounts().

Referenced by FloatHistogram().

void stira::histogram::FloatHistogram::SetEpsilonThreshold ( double epsilonValue )

Sets the threshold value to compare values in the histogram with each other or with zero.

Parameters

epsilonValue the epsilon threshold value

Referenced by FloatHistogram().

bool stira::histogram::FloatHistogram::Write ( std::string fileName )

virtual

writes values from current histogram to a file

Parameters

fileName name of file to write values in

Implements stira::histogram::HistogramMaster< double >.

Referenced by stira::filter::LocalBinaryPattern::ComputeJensenShannonDivergenceBetweenImagePatches(), and FloatHistogram().

The documentation for this class was generated from the following files:

histogram/histogram/FloatHistogram.h
histogram/histogram/FloatHistogram.cpp

Public Member Functions

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation