lti::activeShapeModel Class Reference

This is the base class for active shape models (ASM). More...

#include <ltiActiveShapeModel.h>

Inheritance diagram for lti::activeShapeModel:

[legend]

Collaboration diagram for lti::activeShapeModel:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class activeShapeModel More...
Public Member Functions
	activeShapeModel ()
	activeShapeModel (const activeShapeModel &other)
virtual	~activeShapeModel ()
virtual const char *	getTypeName () const
activeShapeModel &	copy (const activeShapeModel &other)
activeShapeModel &	operator= (const activeShapeModel &other)
virtual functor *	clone () const
const parameters &	getParameters () const
Transformation
Methods for transforming a shape from shape space to eigenspace and vice versa. Eigenspace (w) -> shape space (x): x = eigenVectorMatrix * weightsVector + meanShape Shape space (x) -> eigenspace (w): w = (eigenVectorMatrix)^T * (shape - meanShape)
void	convertShapeToWeightsVector (const pointDistributionModel::shape &src, vector< double > &dest) const
void	convertShapeToWeightsVector (const pointDistributionModel::shape &src, vector< double > &dest, const fvector &pointReliability) const
void	convertWeightsVectorToShape (const vector< double > &src, pointDistributionModel::shape &dest) const
Comparison
Methods for comparing and aligning two shapes.
geometricTransform::parameters	alignShape (pointDistributionModel::shape &shape, const pointDistributionModel::shape &referenceShape) const
geometricTransform::parameters	alignShape (pointDistributionModel::shape &shape, const pointDistributionModel::shape &referenceShape, const fvector &pointReliability) const
float	calculateShapeDifference (const pointDistributionModel::shape &firstShape, const pointDistributionModel::shape &secondShape) const
float	calculateShapeDifference (const pointDistributionModel::shape &firstShape, const pointDistributionModel::shape &secondShape, const fvector &pointReliability) const
Properties
Methods for computing properties of one shape.
tpoint< float >	calculateShapeCenter (const pointDistributionModel::shape &shape) const
tpoint< float >	calculateShapeCenter (const pointDistributionModel::shape &shape, const fvector &pointReliability) const
float	calculateShapeDistance (const pointDistributionModel::shape &shape, const tpoint< float > &thePoint=tpoint< float >(0.0f, 0.0f)) const
float	calculateShapeDistance (const pointDistributionModel::shape &shape, const fvector &pointReliability, const tpoint< float > &thePoint=tpoint< float >(0.0f, 0.0f)) const
double	calculateShapeOrientation (const pointDistributionModel::shape &shape, const pointDistributionModel::shape &referenceShape, const tpoint< float > &theCenter=tpoint< float >(0.0f, 0.0f)) const
double	calculateShapeOrientation (const pointDistributionModel::shape &shape, const pointDistributionModel::shape &referenceShape, const fvector &pointReliability, const tpoint< float > &theCenter=tpoint< float >(0.0, 0.0)) const
Protected Member Functions
float	calculateRelativeShapeScale (const pointDistributionModel::shape &shape, const pointDistributionModel::shape &referenceShape, const fvector &pointReliability, const tpoint< float > &shapeCenter=tpoint< float >(0.0, 0.0), const tpoint< float > &referenceShapeCenter=tpoint< float >(0.0, 0.0)) const
void	trimShape (pointDistributionModel::shape &srcdest) const
void	trimShape (pointDistributionModel::shape &srcdest, const fvector &pointReliability) const
void	trimWeightsVector (vector< double > &srcdest) const
double	calculateAngle (const tpoint< float > &thePoint) const
vector< double >	buildVector (const pointDistributionModel::shape &src) const
pointDistributionModel::shape	buildShape (const vector< double > &src) const
void	createListOfNormals (const pointDistributionModel::shape &theShape, tpointList< float > &listOfNormals) const

Detailed Description

This is the base class for active shape models (ASM).

ASMs are deformable 2D-templates, that can align themselves to image features according to defined search strategies (see derived classes). The deformation is regularized (i.e. constrained) by an underlying PDM (pointDistributionModel).

This class provides the methods for aligning PDM shapes (pointDistributionModel::shape) and to correct invalid shapes to satisfy the restrictions of a given PDM. Furthermore, methods to obtain information about the position of a shape are provided. Shapes with weighted points (point reliabilities) are also supported.

Constructor & Destructor Documentation

lti::activeShapeModel::activeShapeModel ( )

default constructor

lti::activeShapeModel::activeShapeModel ( const activeShapeModel & other )

copy constructor

Parameters:

other

the object to be copied

virtual lti::activeShapeModel::~activeShapeModel ( ) [virtual]

destructor

Member Function Documentation

geometricTransform::parameters lti::activeShapeModel::alignShape	(	pointDistributionModel::shape &	shape,
		const pointDistributionModel::shape &	referenceShape,
		const fvector &	pointReliability
	)			const

align a given weighted shape to another.

Parameters:

	shape	shape to be aligned. the transformed shape will also be saved here.
	referenceShape	fixed shape to be aligned to
	pointReliability	vector containing reliability values between 0 (unreliable) and 1 (reliable) for all points

Returns:: geometric parameters that are used to re-transform shape

geometricTransform::parameters lti::activeShapeModel::alignShape	(	pointDistributionModel::shape &	shape,
		const pointDistributionModel::shape &	referenceShape
	)			const

align a given shape to another.

note that the return geometricTransform::parameters describe the transformation from the referenceShape to the unaligned shape! this is done to align a given shape to a normalized shape and be able to easily project this normalized shape (e.g. meanShape) on the original shape.

Parameters:

	shape	shape to be aligned. the transformed shape will also be saved here.
	referenceShape	fixed shape to be aligned to

Returns:: geometric parameters that are used to re-transform shape

pointDistributionModel::shape lti::activeShapeModel::buildShape ( const vector< double > & src ) const [protected]

convert a vector [x1, y1, x2, y2, .

..] into a shape [(x1,y1),...]

vector<double> lti::activeShapeModel::buildVector ( const pointDistributionModel::shape & src ) const [protected]

convert the shape [(x1,y1),.

..] to a vector [x1, y1, x2, y2, ...]

double lti::activeShapeModel::calculateAngle ( const tpoint< float > & thePoint ) const [protected]

calculate angle between a point and the x-axis

float lti::activeShapeModel::calculateRelativeShapeScale	(	const pointDistributionModel::shape &	shape,
		const pointDistributionModel::shape &	referenceShape,
		const fvector &	pointReliability,
		const tpoint< float > &	shapeCenter = `tpoint< float >(0.0, 0.0)`,
		const tpoint< float > &	referenceShapeCenter = `tpoint< float >(0.0, 0.0)`
	)			const `[protected]`

calculate reliability-weighted scaling factor

tpoint<float> lti::activeShapeModel::calculateShapeCenter	(	const pointDistributionModel::shape &	shape,
		const fvector &	pointReliability
	)			const

calculate the center of a given shape.

all points of the shape are multiplied with their reliability factor, then summed, and the results divided by the sum of reliabilities.

tpoint<float> lti::activeShapeModel::calculateShapeCenter ( const pointDistributionModel::shape & shape ) const

calculate the center of a given shape.

all points of the shape are summed and the results divided by the number of points.

Returns:: center of the given shape

float lti::activeShapeModel::calculateShapeDifference	(	const pointDistributionModel::shape &	firstShape,
		const pointDistributionModel::shape &	secondShape,
		const fvector &	pointReliability
	)			const

calculate difference of two weighted shapes

float lti::activeShapeModel::calculateShapeDifference	(	const pointDistributionModel::shape &	firstShape,
		const pointDistributionModel::shape &	secondShape
	)			const

calculate mean difference of two shapes, i.e.

the sum of squared distances of all corresponding points divided by the number of points

float lti::activeShapeModel::calculateShapeDistance	(	const pointDistributionModel::shape &	shape,
		const fvector &	pointReliability,
		const tpoint< float > &	thePoint = `tpoint< float >(0.0f, 0.0f)`
	)			const

calculate the mean distance of a given point to the reliability-weighted points of a shape

float lti::activeShapeModel::calculateShapeDistance	(	const pointDistributionModel::shape &	shape,
		const tpoint< float > &	thePoint = `tpoint< float >(0.0f, 0.0f)`
	)			const

calculate the mean distance of a given point to the points of a shape

Returns:: average distance

double lti::activeShapeModel::calculateShapeOrientation	(	const pointDistributionModel::shape &	shape,
		const pointDistributionModel::shape &	referenceShape,
		const fvector &	pointReliability,
		const tpoint< float > &	theCenter = `tpoint< float >(0.0, 0.0)`
	)			const

calculate the relative orientation difference of two shapes with regard to a given rotation center and given point reliabilities

double lti::activeShapeModel::calculateShapeOrientation	(	const pointDistributionModel::shape &	shape,
		const pointDistributionModel::shape &	referenceShape,
		const tpoint< float > &	theCenter = `tpoint< float >(0.0f, 0.0f)`
	)			const

calculate the relative orientation difference of two shapes, with regard to a given rotation center

Returns:: angle between referenceShape and shape

virtual functor* lti::activeShapeModel::clone ( ) const [virtual]

returns a pointer to a clone of this functor.

Implements lti::functor.

Reimplemented in lti::gradientASM, and lti::skinASM.

void lti::activeShapeModel::convertShapeToWeightsVector	(	const pointDistributionModel::shape &	src,
		vector< double > &	dest,
		const fvector &	pointReliability
	)			const

convert a reliability-weighted shape to a weights vector, i.e.

project a shape into the pdm eigenspace Formula: w = (eigenVectorMatrix)^T * ((shape - meanShape) * reliabilities) Note: the shape is assumed to be aligned to position, size and rotation of meanShape!

Parameters:

	src	shape with the source data.
	dest	vector where the result will be left.
	pointReliability	vector containing point reliabilities.

void lti::activeShapeModel::convertShapeToWeightsVector	(	const pointDistributionModel::shape &	src,
		vector< double > &	dest
	)			const

convert a shape to a weights vector, i.e.

project a shape into the pdm eigenspace

Formula: w = (eigenVectorMatrix)^T * (shape - meanShape)

Note: the shape is assumed to be aligned to position, size and rotation of meanShape!

Parameters:

	src	shape with the source data.
	dest	vector where the result will be left.

void lti::activeShapeModel::convertWeightsVectorToShape	(	const vector< double > &	src,
		pointDistributionModel::shape &	dest
	)			const

convert a weights vector into a shape (i.e.

shape space) Formula: x = eigenVectorMatrix * weightsVector + meanShape

Parameters:

	src	vector with the source data.
	dest	shape where the result will be left.

activeShapeModel& lti::activeShapeModel::copy ( const activeShapeModel & other )

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:: a reference to this functor object

Reimplemented from lti::functor.

Reimplemented in lti::gradientASM, and lti::skinASM.

void lti::activeShapeModel::createListOfNormals	(	const pointDistributionModel::shape &	theShape,
		tpointList< float > &	listOfNormals
	)			const `[protected]`

create normals for each point

const parameters& lti::activeShapeModel::getParameters ( ) const

returns used parameters

Reimplemented from lti::functor.

Reimplemented in lti::gradientASM, and lti::skinASM.

virtual const char* lti::activeShapeModel::getTypeName ( ) const [virtual]

returns the name of this type ("activeShapeModel")

Reimplemented from lti::functor.

Reimplemented in lti::gradientASM, and lti::skinASM.

activeShapeModel& lti::activeShapeModel::operator= ( const activeShapeModel & other )

alias for copy member

Parameters:

other

the functor to be copied

Returns:: a reference to this functor object

Reimplemented from lti::functor.

Reimplemented in lti::gradientASM, and lti::skinASM.

void lti::activeShapeModel::trimShape	(	pointDistributionModel::shape &	srcdest,
		const fvector &	pointReliability
	)			const `[protected]`

convert reliability-weighted shape to weights vector, correct if invalid and convert back to shape space.

Parameters:

	srcdest	shape with the source data. The result will be left here too.
	pointReliability	the reliabilities of each point of the shape as value between 0.0 and 1.0

void lti::activeShapeModel::trimShape ( pointDistributionModel::shape & srcdest ) const [protected]

convert shape to weights vector, correct if invalid and convert back to shape space.

Parameters:

srcdest

shape with the source data. The result will be left here too.

void lti::activeShapeModel::trimWeightsVector ( vector< double > & srcdest ) const [protected]

Correct weights vector according to varianceCoefficient.

all vector entries that are not in between -c*v and c*v, where c is the given varianceCoefficient and v is the variance of the corresponding dimension (eigenvector), are set to 0.0

Parameters:

srcdest

vector<double> with the source data. The result will be left here too.

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

ltiActiveShapeModel.h

lti::activeShapeModel Class Reference

Classes

Public Member Functions

Protected Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation