This class computes a run length matrix (histogram) from a given image and a mask image if provided. Run length matrces are used for image texture description. More...

#include <itkEnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter.h>

Inheritance diagram for itk::Statistics::EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter< TImageType, THistogramFrequencyContainer >:

[legend]

Collaboration diagram for itk::Statistics::EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter< TImageType, THistogramFrequencyContainer >:

[legend]

Public Types
typedef EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter	Self

typedef ProcessObject	Superclass

typedef SmartPointer< Self >	Pointer

typedef SmartPointer< const Self >	ConstPointer

typedef TImageType	ImageType

typedef ImageType::Pointer	ImagePointer

typedef ImageType::ConstPointer	ImageConstPointer

typedef ImageType::PixelType	PixelType

typedef ImageType::IndexType	IndexType

typedef ImageType::RegionType	RegionType

typedef ImageType::SizeType	RadiusType

typedef ImageType::OffsetType	OffsetType

typedef VectorContainer< unsigned char, OffsetType >	OffsetVector

typedef OffsetVector::Pointer	OffsetVectorPointer

typedef ImageType::PointType	PointType

typedef NumericTraits< PixelType >::RealType	MeasurementType

typedef NumericTraits< PixelType >::RealType	RealType

typedef Histogram< MeasurementType, THistogramFrequencyContainer >	HistogramType

typedef HistogramType::Pointer	HistogramPointer

typedef HistogramType::ConstPointer	HistogramConstPointer

typedef HistogramType::MeasurementVectorType	MeasurementVectorType

Public Member Functions
virtual const char *	GetClassName () const

	itkStaticConstMacro (ImageDimension, unsigned int, TImageType::ImageDimension)

	itkStaticConstMacro (DefaultBinsPerAxis, unsigned int, 256)

virtual void	SetOffsets (OffsetVector *_arg)

void	SetOffset (const OffsetType offset)

void	AddOffsets (const std::vector< OffsetType > offset)

	itkGetModifiableObjectMacro (Offsets, OffsetVector)

virtual void	SetNumberOfBinsPerAxis (unsigned int _arg)

virtual unsigned int	GetNumberOfBinsPerAxis () const

void	SetPixelValueMinMax (PixelType min, PixelType max)

virtual PixelType	GetMin () const

virtual PixelType	GetMax () const

void	SetDistanceValueMinMax (RealType min, RealType max)

virtual RealType	GetMinDistance () const

virtual RealType	GetMaxDistance () const

void	SetInput (const ImageType *image)

const ImageType *	GetInput () const

void	SetMaskImage (const ImageType *image)

const ImageType *	GetMaskImage () const

const HistogramType *	GetOutput () const

double *	GetSiMatrix () const

virtual void	SetInsidePixelValue (PixelType _arg)

virtual PixelType	GetInsidePixelValue () const

Static Public Member Functions
static Pointer	New ()

Protected Types
typedef DataObject::Pointer	DataObjectPointer

typedef ProcessObject::DataObjectPointerArraySizeType	DataObjectPointerArraySizeType

Protected Member Functions
	EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter ()

	~EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter () override

void	PrintSelf (std::ostream &os, Indent indent) const ITK_OVERRIDE

DataObjectPointer	MakeOutput (DataObjectPointerArraySizeType idx) ITK_OVERRIDE

void	GenerateData () ITK_OVERRIDE

void	NormalizeOffsetDirection (OffsetType &offset)

Detailed Description

template<typename TImageType, typename THistogramFrequencyContainer = DenseFrequencyContainer2>
class itk::Statistics::EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter< TImageType, THistogramFrequencyContainer >

This class computes a run length matrix (histogram) from a given image and a mask image if provided. Run length matrces are used for image texture description.

This filters creates a grey-level run length matrix from a N-D scalar image. This is another possible texture description. See the following references. M. M. Galloway. Texture analysis using gray level run lengths. Computer Graphics and Image Processing, 4:172-179, 1975.

A. Chu, C. M. Sehgal, and J. F. Greenleaf. Use of gray value distribution of run lengths for texture analysis. Pattern Recognition Letters, 11:415-420, 1990.

B. R. Dasarathy and E. B. Holder. Image characterizations based on joint gray-level run-length distributions. Pattern Recognition Letters, 12:490-502, 1991.

The basic idea is as follows: Given an image and an offset (e.g. (1, -1) for a 2-d image), each element in the joint histogram describes the frequency for a particular distance/ intensity pair within a given image. This distance/intensity pair can be described as follows: we start at a given voxel which has some intensity. We then "jump" to neighboring pixels in increments provided by the offset(s) as long as the pixel to which we are jumping is within the same intensity bin as the original voxel. The distance component is given by the distance from the original to the final voxel satisfying our jumping criteria.

The offset (or offsets) along which the co-occurences are calculated can be set by the user. Traditionally, only one offset is used per histogram, and offset components in the range [-1, 1] are used. For rotation-invariant features averages of features computed over several histograms with different offsets are generally used, instead of computing features from one histogram create with several offsets. Additionally, instead of using offsets of two or more pixels in any direction, multi-resolution techniques (e.g. image pyramids) are generally used to deal with texture at different spatial resolutions.

This class calculates a 2-d histogram of all the intensity/distance pairs in the given image's requested region, for a given set of offsets. That is, if a given offset falls outside of the requested region (or outside the mask) at a particular point, that distance/intensity pair will not be added to the matrix.

The number of histogram bins on each axis can be set (defaults to 256). Also, by default the histogram min and max corresponds to the largest and smallest possible pixel value of that pixel type. To customize the histogram bounds for a given image, the max and min pixel values that will be placed in the histogram can be set manually. NB: The min and max are INCLUSIVE.

Further, the type of histogram frequency container used is an optional template parameter. By default, a dense container is used, but for images with little texture or in cases where the user wants more histogram bins, a sparse container can be used for the histogram instead.

WARNING: This probably won't work for pixels of double or long-double type unless you set the histogram min and max manually. This is because the largest histogram bin by default has max value of the largest possible pixel value plus 1. For double and long-double types, whose "RealType" as defined by the NumericTraits class is the same, and thus cannot hold any larger values, this would cause a float overflow.

IJ article: https://hdl.handle.net/1926/1374

See also: ScalarImageToNeighbourhoodGreyLevelDifferenceFeaturesFilter; EnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter; HistogramToNeighbourhoodGreyLevelDifferenceFeaturesFilter

Author: : Nick Tustison

Definition at line 117 of file itkEnhancedScalarImageToNeighbourhoodGreyLevelDifferenceMatrixFilter.h.