lti::loweGradientFeature Class Reference

Local feature based on the gradient distribution at the level of the location. More...

#include <ltiLoweGradientFeature.h>

Inheritance diagram for lti::loweGradientFeature:

[legend]

Collaboration diagram for lti::loweGradientFeature:

[legend]

List of all members.

Classes
class	parameters
	The parameters for the class loweGradientFeature. More...
Public Member Functions
	loweGradientFeature ()
	loweGradientFeature (const parameters &par)
	loweGradientFeature (const loweGradientFeature &other)
virtual	~loweGradientFeature ()
virtual const char *	getTypeName () const
loweGradientFeature &	copy (const loweGradientFeature &other)
loweGradientFeature &	operator= (const loweGradientFeature &other)
virtual functor *	clone () const
const parameters &	getParameters () const
virtual bool	updateParameters ()
bool	apply (const channel &src, const std::list< location > &locs, std::list< dvector > &vcts)
bool	apply (const image &src, const std::list< location > &locs, std::list< dvector > &vcts)
bool	apply (const std::list< location > &locs, std::list< dvector > &vcts)
bool	apply (const location &locs, dvector &dest)
bool	generate (const image &src)
bool	generate (const channel &src)
Debug methods.

const imatrix &	getHistogramMask () const
const matrix< float > &	getWeightMask () const
const scaleSpacePyramid < channel > &	getMagnitudePyramid () const
const scaleSpacePyramid < channel > &	getAnglePyramid () const
Protected Member Functions
bool	computeHistograms (const int level, const tpoint< float > &pos, const float angle, matrix< double > &hists) const
Protected Attributes
scaleSpacePyramid< channel >	pyrMag
LUT and Masks

matrix< int >	hmask
matrix< float >	weight
tpoint< float >	maskCenter
int	numHistograms
matrix< int >	ioPts
bool	initMasks ()

Detailed Description

Local feature based on the gradient distribution at the level of the location.

This local descriptor based was developed by David Lowe (see his paper "Distinctive Image Features from Scale-Invariant Keypoints", 2003).

Lowe proposed also proposed 1999 a similar local descriptor called "SIFT", but he changed it in his later work to used the information of one resolution only.

It computes histograms of the gradient orientation for several parts of the region delimited by a location, where the weights of each sample are determined by the magnitude of the gradient and the distance to the center of the location.

The location is divided in each axis by a given integer number h (parameters::numHistograms). This results in a total of h x h histograms and each one of them will have n x n samples, where n represents parameters::sampleRegionSize.

It can compute features for monochrome images and for color images, where the color feature is just the concatenation of the three monochromatic RGB ones.

Constructor & Destructor Documentation

lti::loweGradientFeature::loweGradientFeature ( )

Default constructor.

lti::loweGradientFeature::loweGradientFeature ( const parameters & par )

Construct a functor using the given parameters.

lti::loweGradientFeature::loweGradientFeature ( const loweGradientFeature & other )

Copy constructor.

Parameters:

other

the object to be copied

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

Destructor.

Member Function Documentation

bool lti::loweGradientFeature::apply	(	const location &	locs,
		dvector &	dest
	)

Compute the gradient feature for the given location.

You must call first the method generate() to indicate which image needs to be analyzed. (Otherwise it would take too much time if you need to extract several features from the same image, and the pyramid would have to be generated each time!)

Parameters:

	locs	location that need to be analyzed.
	dest	feature vector.

Returns:: true if successful, or false otherwise.

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

Extract features based on a previously generated pyramid set.

The decision if you want color or gray features is taken at the time of the pyramid generation.

See also:: generate()

bool lti::loweGradientFeature::apply	(	const image &	src,
		const std::list< location > &	locs,
		std::list< dvector > &	vcts
	)

Extracts the color features at the loctions.

bool lti::loweGradientFeature::apply	(	const channel &	src,
		const std::list< location > &	locs,
		std::list< dvector > &	vcts
	)

Extracts gradient features at the locations for a gray-valued input channel.

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

Returns a pointer to a clone of this functor.

Reimplemented from lti::featureExtractor.

bool lti::loweGradientFeature::computeHistograms	(	const int	level,
		const tpoint< float > &	pos,
		const float	angle,
		matrix< double > &	hists
	)			const `[protected]`