lti::schieleCrowley6DFeature Class Reference

This functor create huge lists of local features, that can be used to train the multidimensional receptive field histograms of Schiele and Crowley (see lti::shClassifier). More...

#include <ltiSchieleCrowley6DFeature.h>

Inheritance diagram for lti::schieleCrowley6DFeature:

[legend]

Collaboration diagram for lti::schieleCrowley6DFeature:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class schieleCrowley6DFeature More...
Public Member Functions
	schieleCrowley6DFeature ()
	schieleCrowley6DFeature (const schieleCrowley6DFeature &other)
virtual	~schieleCrowley6DFeature ()
virtual const char *	getTypeName () const
schieleCrowley6DFeature &	copy (const schieleCrowley6DFeature &other)
schieleCrowley6DFeature &	operator= (const schieleCrowley6DFeature &other)
virtual functor *	clone () const
const parameters &	getParameters () const
slow apply methods
Use these methods only when you need the src channel once. To extract feature for different regions of the same src channel, see for example apply(channel8&,std::list<dvector>&)
bool	apply (const channel &src, const channel8 &mask, std::list< dvector > &vcts) const
bool	apply (const channel &src, const areaPoints &mask, std::list< dvector > &vcts) const
bool	apply (const channel &src, const std::list< location > &mask, std::list< dvector > &vcts) const
bool	apply (const channel &src, const double angle, const channel8 &mask, std::list< dvector > &vcts) const
bool	apply (const channel &src, const double angle, const areaPoints &mask, std::list< dvector > &vcts) const
bool	apply (const channel &src, const double angle, const std::list< location > &mask, std::list< dvector > &vcts) const
fast apply methods.
All these apply methods use the internal computed gaussian derivatives. Before calling these methods you need therefore to specify your channel with the method use(). This approach is very fast when extracting different feature sets for the same image (for example, feature sets for each object in the image) In this case you specify once with use() which channel you want to use, and after that, you can call the apply method for each object found in the image. This way, the expensive filters will be applied just one, and the feature extraction is faster.
bool	use (const channel &chnl)
bool	apply (const channel8 &mask, std::list< dvector > &vcts) const
bool	apply (const areaPoints &mask, std::list< dvector > &vcts) const
bool	apply (const std::list< location > &mask, std::list< dvector > &vcts) const
bool	apply (const double angle, const channel8 &mask, std::list< dvector > &vcts) const
bool	apply (const double angle, const areaPoints &mask, std::list< dvector > &vcts) const
bool	apply (const double angle, const std::list< location > &mask, std::list< dvector > &vcts) const
Protected Member Functions
void	getBasisChannels (const channel &chnl, const double sigma, channel &Dx, channel &Dy) const
Protected Attributes
channel *	chnls

Detailed Description

This functor create huge lists of local features, that can be used to train the multidimensional receptive field histograms of Schiele and Crowley (see lti::shClassifier).

Each vector will be usually 6 dimensional, where the even dimensions will correspond to the pixel values of gaussian derivatives in x-direction and the odd dimensions will correspond to the y-direction.

Each pair of dimensions will correspond to one resolution or scale, where the std. deviation of the gaussians between the levels will be related by a factor 2.

More information on the features and classification approach can be taken from the original paper:

B. Schiele and J. Crowley. Recognition without correspondence using multidimensional receptive field histograms. Technical Report No. 453 MIT Medial Laboratory, Perceptial Computing Section.

Constructor & Destructor Documentation

lti::schieleCrowley6DFeature::schieleCrowley6DFeature ( )

default constructor

lti::schieleCrowley6DFeature::schieleCrowley6DFeature ( const schieleCrowley6DFeature & other )

copy constructor

Parameters:

other

the object to be copied

virtual lti::schieleCrowley6DFeature::~schieleCrowley6DFeature ( ) [virtual]

destructor

Member Function Documentation

bool lti::schieleCrowley6DFeature::apply	(	const double	angle,
		const std::list< location > &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	angle	reorient the gaussian derivatives with this angle
	mask	only the central pixels of the locations will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const double	angle,
		const areaPoints &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	angle	reorient the gaussian derivatives with this angle
	mask	only the pixels in the mask (areaPoints list) will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const double	angle,
		const channel8 &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	angle	reorient the gaussian derivatives with this angle
	mask	only the pixels in src, for which the respective pixel in the mask is different than zero will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const std::list< location > &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	mask	only the central pixels in the locations will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const areaPoints &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	mask	only the pixels in the mask (areaPoints list) will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel8 &	mask,
		std::list< dvector > &	vcts
	)			const

Before calling this method, you need to specify your channel with use(const channel&).

Parameters:

	mask	only the pixels in src, for which the respective pixel in the mask is different than zero will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const double	angle,
		const std::list< location > &	mask,
		std::list< dvector > &	vcts
	)			const

Get a vector for each location in the locations list.

Parameters:

	src	original channel with the source data
	angle	reorient the gaussian derivatives with this angle
	mask	only the central pixels of the locations will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const double	angle,
		const areaPoints &	mask,
		std::list< dvector > &	vcts
	)			const

Get a vector for each pixel in the areaPoint.

Parameters:

	src	original channel with the source data
	angle	reorient the gaussian derivatives with this angle
	mask	only the pixels in the mask (areaPoints list) will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const double	angle,
		const channel8 &	mask,
		std::list< dvector > &	vcts
	)			const

Get a vector for each non-zero pixel in the mask.

Parameters:

	src	original channel with the source data
	angle	reorient the gaussian derivatives with this angle
	mask	only the pixels in src, for which the respective pixel in the mask is different than zero will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const std::list< location > &	mask,
		std::list< dvector > &	vcts
	)			const `[virtual]`

Get a vector for each location in the locations list.

Parameters:

	src	original channel with the source data
	mask	only the central pixels in the locations will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

Reimplemented from lti::localFeatureExtractor.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const areaPoints &	mask,
		std::list< dvector > &	vcts
	)			const

Get a vector for all pixels in the area points.

Parameters:

	src	original channel with the source data
	mask	only the pixels in the mask (areaPoints list) will be considered.
	vcts	the list of all feature vectors for all masked pixels

Returns:: true if apply successful or false otherwise.

bool lti::schieleCrowley6DFeature::apply	(	const channel &	src,
		const channel8 &	mask,
		std::list< dvector > &	vcts
	)			const

Get a vector for all non-zero pixels in the mask, using the gaussian derivatives of the given channel.

Parameters:

	src	original channel with the source data
	mask	only the pixels in src, for which the respective pixel in the mask is different than zero will be considered.
	vcts	the list of all feature vectors for all masked pixels