#include <itkImagePCAShapeModelEstimator.h>

Inheritance diagram for itk::ImagePCAShapeModelEstimator< TInputImage, TOutputImage >:

Collaboration diagram for itk::ImagePCAShapeModelEstimator< TInputImage, TOutputImage >:

Public Types
typedef SmartPointer< const Self >	ConstPointer
typedef ImageRegionConstIterator < TInputImage >	InputImageConstIterator
typedef TInputImage::ConstPointer	InputImageConstPointer
typedef ImageRegionIterator < TInputImage >	InputImageIterator
typedef TInputImage::PixelType	InputImagePixelType
typedef TInputImage::Pointer	InputImagePointer
typedef TInputImage	InputImageType
typedef vnl_matrix< double >	MatrixOfDoubleType
typedef vnl_matrix< int >	MatrixOfIntegerType
typedef ImageRegionIterator < TOutputImage >	OutputImageIterator
typedef TOutputImage::Pointer	OutputImagePointer
typedef TOutputImage	OutputImageType
typedef SmartPointer< Self >	Pointer
typedef ImagePCAShapeModelEstimator	Self
typedef ImageShapeModelEstimatorBase < TInputImage, TOutputImage >	Superclass
typedef vnl_vector< double >	VectorOfDoubleType
Public Member Functions
virtual ::itk::LightObject::Pointer	CreateAnother (void) const
virtual const char *	GetNameOfClass () const
virtual unsigned int	GetNumberOfPrincipalComponentsRequired () const
virtual unsigned int	GetNumberOfTrainingImages () const
virtual void	SetNumberOfPrincipalComponentsRequired (unsigned int n)
virtual void	SetNumberOfTrainingImages (unsigned int n)
Static Public Member Functions
static Pointer	New ()
Static Public Attributes
static const unsigned int	InputImageDimension = TInputImage::ImageDimension
Protected Member Functions
virtual void	EnlargeOutputRequestedRegion (DataObject *)
void	GenerateData ()
virtual void	GenerateInputRequestedRegion ()
Private Types
typedef TInputImage::SizeType	ImageSizeType
typedef std::vector < InputImageConstIterator >	InputImageIteratorArray
typedef std::vector < InputImageConstPointer >	InputImagePointerArray
typedef TInputImage::PixelType	InputPixelType
Private Member Functions
void	CalculateInnerProduct ()
void	EstimatePCAShapeModelParameters ()
virtual void	EstimateShapeModels ()
	ImagePCAShapeModelEstimator (const Self &)
void	operator= (const Self &)
Private Attributes
VectorOfDoubleType	m_EigenValues
VectorOfDoubleType	m_EigenVectorNormalizedEnergy
MatrixOfDoubleType	m_EigenVectors
MatrixOfDoubleType	m_InnerProduct
InputImageIteratorArray	m_InputImageIteratorArray
ImageSizeType	m_InputImageSize
VectorOfDoubleType	m_Means
unsigned int	m_NumberOfPixels
unsigned int	m_NumberOfPrincipalComponentsRequired
unsigned int	m_NumberOfTrainingImages
virtual VectorOfDoubleType	GetEigenValues () const
	ImagePCAShapeModelEstimator ()
	~ImagePCAShapeModelEstimator ()
virtual void	PrintSelf (std::ostream &os, Indent indent) const

Detailed Description

template<class TInputImage, class TOutputImage = Image< double, ::itk::GetImageDimension< TInputImage >::ImageDimension >>
class itk::ImagePCAShapeModelEstimator< TInputImage, TOutputImage >

Base class for ImagePCAShapeModelEstimator object.

itkImagePCAShapeModelEstimator performs a principal component analysis (PCA) on a set of images. The user specifies the number of training images and also the number of desired largest principal components needed. The ITK pipeline mechanism sets up the storage for both input and output images. The number of output images are the user specified number of desired largest principal components plus 1 (for the mean image).

The algorithm uses the VNL library to perform the eigen analysis. To speed the computation of the instead of performing the eigen analysis of the covariance vector A*A' where A is a matrix with p x t, p = number of pixels or voxels in each images and t = number of training images, we calculate the eigen vectors of the inner product matrix A'*A. The resulting eigen vectors (E) are then multiplied with the the matrix A to get the principal compoenets. The covariance matrix has a dimension of p x p. Since number of pixels in any image being typically very high the eigen decomposition becomes computationally expensive. The inner product on the other hand has the dimension of t x t, where t is typically much smaller that p. Hence the eigen decomposition (most compute intensive part) is an orders of magnitude faster.

The Update() function enables the calculation of the various models, creates the membership function objects and populates them.

Wiki Examples:

All Examples

Compute a PCA shape model from a training sample

Definition at line 77 of file itkImagePCAShapeModelEstimator.h.