This class takes scalar image as input, and outputs a vector field of image gradient. More...

Inheritance diagram for itk::ScalarImageToNormalizedGradientVectorImageFilter< TInputImage, TScalarType >:

Collaboration diagram for itk::ScalarImageToNormalizedGradientVectorImageFilter< TInputImage, TScalarType >:

Public Types
enum	DerivativeType { CENTRAL_DIFFERENCES, DERIVATIVE_OPERATOR, DERIVATIVE_OF_GAUSSIAN }

typedef TScalarType	VectorDataType

typedef ScalarImageToNormalizedGradientVectorImageFilter	Self

typedef ImageToImageFilter < TInputImage, Image< Vector < VectorDataType, TInputImage::ImageDimension > , TInputImage::ImageDimension > >	Superclass

typedef SmartPointer< Self >	Pointer

typedef SmartPointer< const Self >	ConstPointer

typedef Vector< VectorDataType, itkGetStaticConstMacro(Dimension) >	OutputPixelType

typedef CovariantVector < VectorDataType, itkGetStaticConstMacro(Dimension) >	CovariantVectorType

typedef Image< OutputPixelType, itkGetStaticConstMacro(Dimension) >	OutputImageType

typedef Image < CovariantVectorType, itkGetStaticConstMacro(Dimension) >	CovariantVectorImageType

typedef CovariantVectorImageType::Pointer	CovariantVectorImagePointer

typedef Superclass::InputImageType	InputImageType

typedef InputImageType::PixelType	InputPixelType

typedef InputImageType::RegionType	InputRegionType

typedef GradientImageFilter < TInputImage, VectorDataType, VectorDataType >	GradientImageFilterType

typedef GradientImageFilterType::Pointer	GradientImageFilterPointer

typedef GradientRecursiveGaussianImageFilter < TInputImage, CovariantVectorImageType >	GradientRecursiveGaussianImageFilterType

typedef GradientRecursiveGaussianImageFilterType::Pointer	GradientRecursiveGaussianImageFilterPointer

typedef GaussianSmoothVectorFieldFilter < VectorDataType, itkGetStaticConstMacro(Dimension), itkGetStaticConstMacro(Dimension)>	GaussianSmoothFilterType

typedef GaussianSmoothFilterType::Pointer	GaussianSmoothFilterPointer

typedef NormaliseVectorFilter < VectorDataType, itkGetStaticConstMacro(Dimension) >	NormaliseFilterType

typedef NormaliseFilterType::Pointer	NormaliseFilterPointer

Public Member Functions
virtual const char *	GetClassName () const

	itkStaticConstMacro (Dimension, unsigned int, TInputImage::ImageDimension)

void	SetScalarImage (const InputImageType *image)

virtual void	SetUseMillimetreScaling (bool _arg)

virtual bool	GetUseMillimetreScaling ()

virtual void	SetDivideByTwo (bool _arg)

virtual bool	GetDivideByTwo ()

virtual void	SetNormalize (bool _arg)

virtual bool	GetNormalize ()

virtual void	SetPadValue (InputPixelType _arg)

virtual InputPixelType	GetPadValue ()

virtual void	SetDerivativeType (DerivativeType _arg)

virtual DerivativeType	GetDerivativeType ()

virtual void	SetSigma (double _arg)

virtual double	GetSigma ()

Static Public Member Functions
static Pointer	New ()

Protected Member Functions
	ScalarImageToNormalizedGradientVectorImageFilter ()

	~ScalarImageToNormalizedGradientVectorImageFilter ()

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

virtual void	GenerateData ()

Detailed Description

template<class TInputImage, typename TScalarType>
class itk::ScalarImageToNormalizedGradientVectorImageFilter< TInputImage, TScalarType >

This class takes scalar image as input, and outputs a vector field of image gradient.

This class implements step (8) in in Jones et al. Human Brain Mapping 11:12-32 (2000). The input should be the output of itkLaplacianSolverImageFilter, but the filter will work for any scalar image. We simply compute the derivative using 2 point central differences, and normalize it to unit length. The output vector length takes into account anisotropic voxels, so the output vectors are in mm space, and hence are 1mm long.

Additionally, we added a member variable UseMillimetreScaling, so that if true, the size in millimetre of the voxel is taken into account. Defaults to on.

Additionally, we added a member variable, DivideByTwo, so that if true, the difference of the pixel + and pixel - is divided by two. Defaults to on.

Additionally, we added member variable Normalize, which is a boolean to turn normalization to unit length on/off. Defaults to on.

Additionally, we added a pad value. If the current pixel, or +/- 1 pixel in any dimension is == pad value, then the gradient is ignored.

As of 25th Nov 2009, we have wrapped the classes GradientImageFilter and GradientRecursiveGaussianImageFilter.

As of 1st Dec 2009, I have also wrapped smoothing of the vector field. You can supply a sigma (default 0 - i.e. no smoothing) and the resultant vectors will be smoothed. Also, if normalisation is on, the the smoothed vectors come out, normalised to unit length.

See also: LaplacianSolverImageFilter; RelaxStreamlinesFilter; IntegrateStreamlinesFilter; LagrangianInitializedRelaxStreamlinesFilter