lti::gradientASM Class Reference

This class implements an alignment strategy for activeShapeModel (ASM). More...

#include <ltiGradientASM.h>

Inheritance diagram for lti::gradientASM:

[legend]

Collaboration diagram for lti::gradientASM:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class gradientASM More...
Public Member Functions
	gradientASM ()
	gradientASM (const gradientASM &other)
virtual	~gradientASM ()
virtual const char *	getTypeName () const
gradientASM &	copy (const gradientASM &other)
gradientASM &	operator= (const gradientASM &other)
virtual functor *	clone () const
const parameters &	getParameters () const
pointDistributionModel::shape &	apply (pointDistributionModel::shape &srcdest, const channel &gradientChannel) const
Protected Member Functions
pointDistributionModel::shape &	adjustShape (pointDistributionModel::shape &srcdest, const channel &gradientChannel, fvector &pointReliability) const
bool	createWeightingKernel (kernel1D< float > &theKernel) const
void	getIntensityArray (const tpoint< float > &start, const tpoint< float > &normal, const channel &chnl, array< float > &values) const
tpoint< int >	roundPoint (const tpoint< float > &floatPoint) const

Detailed Description

This class implements an alignment strategy for activeShapeModel (ASM).

See: Sonka's "Image Processing, Analysis, and Machine Vision", p380ff.

For this purpose, strong gradients perpendicular to the border of a given shape are searched, starting from each of the shapes's landmark points. The best candidate is selected by gradient strength AND distance to the border (see parameters::weightingKernel). The resulting shape's parameters are then adjusted to fit the underlying pointDistributionModel (PDM) in two steps:

geometric tranformation (coarse fit)
deformation (fine fit)

The procedure is repeated, until a given number of iterations is reached.

Constructor & Destructor Documentation

lti::gradientASM::gradientASM ( )

default constructor

lti::gradientASM::gradientASM ( const gradientASM & other )

copy constructor

Parameters:

other

the object to be copied

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

destructor

Member Function Documentation

pointDistributionModel::shape& lti::gradientASM::adjustShape	(	pointDistributionModel::shape &	srcdest,
		const channel &	gradientChannel,
		fvector &	pointReliability
	)			const `[protected]`

align shape on the given gradient channel (no iterations!)

Returns:: a reference to the srcdest.

pointDistributionModel::shape& lti::gradientASM::apply	(	pointDistributionModel::shape &	srcdest,
		const channel &	gradientChannel
	)			const

align shape on the given gradient channel.

iteratively, scan for closest border: iteration process: the channel's values along a control point's normal are saved and afterwards weighted with the weighting function specified in the parameters class. the control point is then moved to the point with the highest value. the shape is then trimmed to a valid shape.

Parameters:

	srcdest	shape with the source data. The result will be left here too.
	gradientChannel	gradient channel. the pdmShape is aligned along higher values in this channel.

Returns:: a reference to the srcdest.

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

returns a pointer to a clone of this functor.

Reimplemented from lti::activeShapeModel.

Reimplemented in lti::skinASM.

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

copy data of "other" functor.

Parameters:

other

the functor to be copied

Returns:: a reference to this functor object

Reimplemented from lti::activeShapeModel.

Reimplemented in lti::skinASM.

bool lti::gradientASM::createWeightingKernel ( kernel1D< float > & theKernel ) const [protected]

create weighting kernel according to parameters return false, if no kernel was built

void lti::gradientASM::getIntensityArray	(	const tpoint< float > &	start,
		const tpoint< float > &	normal,
		const channel &	chnl,
		array< float > &	values
	)			const `[protected]`