Namespaces
	global
	Global-features extraction.

Classes
class	CCommonFeatureExtractor
	Common class, which unites feature extraction algorithms. More...

class	CCoordinate
	Coordinate feature extraction class. More...

class	CDistance
	Distance feature extraction class. More...

class	CGlobalFeatureExtractor
	Interface class for global feature extraction algorithms. More...

class	CGradient
	Gradient feature extraction class. More...

class	CHOG
	HOG (histogram of oriented gradients) feature extraction class. More...

class	CHSV
	Hue, Saturation and Value feature extraction class. More...

class	CIntensity
	Intensity feature extraction class. More...

class	CNDVI
	NDVI (normalized difference vegetation index) feature extraction class. More...

class	CScale
	Scale feature extraction class. More...

class	CSIFT
	SIFT (scale-invariant feature transform) feature extraction class. More...

class	CSparseCoding
	Sparse Coding feature extraction class. More...

class	CSparseDictionary
	Sparse Dictionary Learning class. More...

class	CVariance
	Variance feature extraction class. More...

class	IFeatureExtractor
	Interface class for feature extraction algorithms. More...

class	ILocalFeatureExtractor
	Interface class for local feature extraction algorithms. More...

struct	SqNeighbourhood
	Square neighborhood structure. More...

Typedefs
typedef struct DirectGraphicalModels::fex::SqNeighbourhood	SqNeighbourhood
	Square neighborhood structure. More...

Enumerations
enum	ChannelsRGB { CH_BLUE, CH_GREEN, CH_RED }
	Channels in the BGR color space. More...

enum	ChannelsHSV { CH_HUE, CH_SATURATION, CH_VALUE }
	Channels in the HSV color space. More...

enum	coordinateType { COORDINATE_ORDINATE, COORDINATE_ABSCISS, COORDINATE_RADIUS }
	Types of the coordinate feature. More...

enum	BasePointLocation { BP_CENTER, BP_LEFT, BP_RIGHT, BP_TOP, BP_BOTTOM }
	Some special cases of the base point location inside the neighborhood. More...

Functions
template<typename T >
T	linear_mapper (float val, float min, float max)
	Linear 1D mapping. More...

template<typename T >
T	two_linear_mapper (float val, float min, float max, float mid, T midPoint)
	Two-linear 1D mapping. More...

SqNeighbourhood	sqNeighbourhood (int leftGap, int rightGap, int upperGap, int lowerGap)
	Initializes the square neighborhood structure. More...

SqNeighbourhood	sqNeighbourhoodAll (int R)
	Initializes the square neighborhood structure with all the same values (base point in the center) More...

SqNeighbourhood	sqNeighbourhood (int R, BasePointLocation location=BP_CENTER)
	Initializes the square neighborhood structure with a pre-define shape. More...

Variables
const float	SC_LRATE_W = 5e-2f
	Learning rate (speed) for weights \(W\). More...

const float	SC_LRATE_D = 1e-2f
	Learning rate (speed) for dictionary \(D\). More...

const float	SC_LAMBDA = 5e-5f
	\(\lambda\): L1-regularisation parameter (on features) More...

const float	SC_EPSILON = 1e-5f
	\(\epsilon\): L1-regularisation epsilon \( \left\\|x\right\\|_1 \approx \sqrt{x^2 + \epsilon} \) More...

const float	SC_GAMMA = 1e-2f
	\(\gamma\): L2-regularisation parameter (on dictionary words) More...

Typedef Documentation

◆ SqNeighbourhood

typedef struct DirectGraphicalModels::fex::SqNeighbourhood DirectGraphicalModels::fex::SqNeighbourhood

Square neighborhood structure.

This structure defines rectangular neighborhood around its base point. The size of the neighborhoood is given via four gap values: \((leftGap+rightGap+1)\times(upperGap+lowerGap+1)\), and its shape is defines as depicted at Figure 1.

This definition of a neighbourhood extends the classical one, where the neighborhood is represented as a square with the base point in the center.

Fig. 1

Enumeration Type Documentation

◆ BasePointLocation

enum DirectGraphicalModels::fex::BasePointLocation

Some special cases of the base point location inside the neighborhood.

Enumerator
BP_CENTER	The base point is located in the neighborhood's center.
BP_LEFT	The base point is located at the neighborhood's left boundry.
BP_RIGHT	The base point is located in the neighborhood's right boundry.
BP_TOP	The base point is located in the neighborhood's upper boundry.
BP_BOTTOM	The base point is located in the neighborhood's lower boundry.

Definition at line 26 of file SquareNeighborhood.h.

◆ ChannelsHSV

enum DirectGraphicalModels::fex::ChannelsHSV

Channels in the HSV color space.

Enumerator
CH_HUE	Hue channel.
CH_SATURATION	Saturation channel.
CH_VALUE	Value channel.

Definition at line 30 of file CommonFeatureExtractor.h.

◆ ChannelsRGB

enum DirectGraphicalModels::fex::ChannelsRGB

Channels in the BGR color space.

Enumerator
CH_BLUE	Blue channel.
CH_GREEN	Green channel.
CH_RED	Red channel.

Definition at line 23 of file CommonFeatureExtractor.h.

◆ coordinateType

enum DirectGraphicalModels::fex::coordinateType

Types of the coordinate feature.

Enumerator
COORDINATE_ORDINATE	Coordinate feature depend on the pixel's ordinate (y-coordinate).
COORDINATE_ABSCISS	Coordinate feature depend on the pixel's absciss (x-coordinate).
COORDINATE_RADIUS	Coordinate feature depend on the pixel's distance to the image center.

Definition at line 12 of file Coordinate.h.

Function Documentation

◆ linear_mapper()

template<typename T >

T DirectGraphicalModels::fex::linear_mapper	(	float	val,
		float	min,
		float	max
	)

inline

Linear 1D mapping.

This function perform linear mapping of the value val from one interval to another: \(val\in[min; max]\rightarrow res\in[T.min; T.max]\), such that:

\begin{eqnarray*} min&\rightarrow&T.min \\ max&\rightarrow&T.max \end{eqnarray*}

Template Parameters

T	The type of the resulting value. Usually byte or word. It also defines the resulting interval, e.g for byte it is [0; 255] and for word it is [0; 65535].

Parameters

val	The value to map.
min	The lower boundary of the val.
max	The higher bounday of the val.

Returns: The mapped value res.

Definition at line 22 of file LinearMapper.h.

◆ sqNeighbourhood() [1/2]

SqNeighbourhood DirectGraphicalModels::fex::sqNeighbourhood	(	int	leftGap,
		int	rightGap,
		int	upperGap,
		int	lowerGap
	)

inline

Initializes the square neighborhood structure.

Parameters

leftGap	Distance from the base point to the neighborhood's left boundary.
rightGap	Distance from the base point to the neighborhood's right boundary.
upperGap	Distance from the base point to the neighborhood's upper boundary.
lowerGap	Distance from the base point to the neighborhood's lower boundary.

Returns: Initialized square neighborhood of the size: \((leftGap+rightGap+1)\times(upperGap+lowerGap+1)\) (Ref. SqNeighbourhood).

Definition at line 42 of file SquareNeighborhood.h.

Here is the caller graph for this function:

◆ sqNeighbourhood() [2/2]

SqNeighbourhood DirectGraphicalModels::fex::sqNeighbourhood	(	int	R,
		BasePointLocation	location = `BP_CENTER`
	)

inline

Initializes the square neighborhood structure with a pre-define shape.

Parameters

R	Distance from the base point to the neighborhood's boundaries
location	Flag describing the location of the base point (Ref. BasePointLocation and Figure 2) Fig. 2

Returns: Initialized square neighborhood of the size \((2R + 1)\times(2R + 1)\) (Ref. SqNeighbourhood)

Definition at line 64 of file SquareNeighborhood.h.

Here is the call graph for this function:

◆ sqNeighbourhoodAll()

SqNeighbourhood DirectGraphicalModels::fex::sqNeighbourhoodAll ( int R )

inline

Initializes the square neighborhood structure with all the same values (base point in the center)

Parameters

R	Distance from the base point to the neighborhood's boundaries (radius)

Returns: Initialized square neighborhood of the size \((2R + 1)\times(2R + 1)\) (Ref. SqNeighbourhood)

Definition at line 56 of file SquareNeighborhood.h.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ two_linear_mapper()

template<typename T >

T DirectGraphicalModels::fex::two_linear_mapper	(	float	val,
		float	min,
		float	max,
		float	mid,
		T	midPoint
	)

inline

Two-linear 1D mapping.

This function perform linear mapping of the value val from one interval to another: \(val\in[min; max]\rightarrow res\in[T.min; T.max]\), such that:

\begin{eqnarray*} min&\rightarrow&T.min \\ mid&\rightarrow&midPoint \\ max&\rightarrow&T.max \end{eqnarray*}

For more detail please refer to the Figure 1.

Fig. 1

Parameters

val	The value to map.
min	The lower boundary of the val.
max	The higher bounday of the val.
mid	The x-coordinate of the intersection point, \(mid\in(min; max)\) (Ref. Figure 1).
midPoint	The y-coordinate of the intersection point, \(midPoint\in[T.min; T.max]\) (Ref. Figure 1).

Returns: The mapped value res.

Definition at line 47 of file LinearMapper.h.

Variable Documentation

◆ SC_EPSILON

const float DirectGraphicalModels::fex::SC_EPSILON = 1e-5f

\(\epsilon\): L1-regularisation epsilon \( \left\|x\right\|_1 \approx \sqrt{x^2 + \epsilon} \)

Definition at line 13 of file SparseDictionary.h.

◆ SC_GAMMA

const float DirectGraphicalModels::fex::SC_GAMMA = 1e-2f

\(\gamma\): L2-regularisation parameter (on dictionary words)

Definition at line 14 of file SparseDictionary.h.

◆ SC_LAMBDA

const float DirectGraphicalModels::fex::SC_LAMBDA = 5e-5f

\(\lambda\): L1-regularisation parameter (on features)

Definition at line 12 of file SparseDictionary.h.

◆ SC_LRATE_D

const float DirectGraphicalModels::fex::SC_LRATE_D = 1e-2f

Learning rate (speed) for dictionary \(D\).

Definition at line 10 of file SparseDictionary.h.

◆ SC_LRATE_W

const float DirectGraphicalModels::fex::SC_LRATE_W = 5e-2f

Learning rate (speed) for weights \(W\).

Definition at line 9 of file SparseDictionary.h.

Namespaces

Classes

Typedefs

Enumerations

Functions

Variables

Typedef Documentation

◆ SqNeighbourhood

Enumeration Type Documentation

◆ BasePointLocation

◆ ChannelsHSV

◆ ChannelsRGB

◆ coordinateType

Function Documentation

◆ linear_mapper()

◆ sqNeighbourhood() [1/2]

◆ sqNeighbourhood() [2/2]

◆ sqNeighbourhoodAll()

◆ two_linear_mapper()

Variable Documentation

◆ SC_EPSILON

◆ SC_GAMMA

◆ SC_LAMBDA

◆ SC_LRATE_D

◆ SC_LRATE_W