lti::colorContrastGradient Class Reference

The contrast gradient is used in color or multi-spectral image as a replacement of the gray-value gradient in gray images. More...

#include <ltiColorContrastGradient.h>

Inheritance diagram for lti::colorContrastGradient:

[legend]

Collaboration diagram for lti::colorContrastGradient:

[legend]

List of all members.

Classes
class	parameters
	the parameters for the class gradientFunctor More...
Public Member Functions
	colorContrastGradient ()
	colorContrastGradient (const parameters::eOutputFormat &format, const int gradKernelSize=0)
	colorContrastGradient (const parameters &par)
	colorContrastGradient (const colorContrastGradient &other)
virtual	~colorContrastGradient ()
virtual const char *	getTypeName () const
bool	apply (const image &src, channel &xOrMag, channel &yOrArg) const
bool	apply (const channel &src, channel &xOrMag, channel &yOrArg) const
bool	apply (const channel8 &src, channel &xOrMag, channel &yOrArg) const
bool	apply (const channel &c1, const channel &c2, const channel &c3, channel &xOrMag, channel &yOrArg) const
bool	apply (const channel &c1, const channel &c2, const channel &c3, channel &xOrMag, channel &yOrArg, float &maxMag) const
bool	apply (const channel &c1, const channel &c2, channel &xOrMag, channel &yOrArg, float &maxMag) const
bool	apply (const channel &c1, const channel &c2, const channel &c3, channel &lambdaMin, channel &lambdaMax, channel &orientation, float &maxMag) const
colorContrastGradient &	copy (const colorContrastGradient &other)
colorContrastGradient &	operator= (const colorContrastGradient &other)
virtual functor *	clone () const
const parameters &	getParameters () const
Protected Member Functions
bool	computeGradientPolar (const channel &c1, const channel &c2, const channel &c3, channel &gradAbs, channel &orientation, float &maxVal) const
bool	computeGradientPolar (const channel &c1, const channel &c2, channel &gradAbs, channel &orientation, float &maxVal) const
bool	computeGradientPolar (const channel &c1, const channel &c2, const channel &c3, channel &lambdaMin, channel &lambdaMax, channel &orientation, float &maxVal) const
bool	computeColorGradientCart (const channel &c1, const channel &c2, channel &xchnl, channel &ychnl, float &maxVal) const
bool	computeColorGradientCart (const channel &c1, const channel &c2, const channel &c3, channel &xchnl, channel &ychnl, float &maxVal) const
float	magSqr (const float a, const float b) const

Detailed Description

The contrast gradient is used in color or multi-spectral image as a replacement of the gray-value gradient in gray images.

Here, the approach introduced in A. Cumani, "Edge Detection in Multispectral Images", Technical Report, Istituto Elettrotecnico Nazionale "Galileo Ferraris", 1989 is followed.

A more detail theoretical analysis can be found in I.R. Greenshields "Coherent computation of the multispectral maximal directional derivative" Image and Vision Computing, Vol 18. 1999, pp. 1-7

With this approach, instead of a gradient, the maximum of the contrast function for each pixel is searched. The contrast function defines the direction at the (x,y) plane at which the contrast change is maximal. It corresponds to the greater eigenvalue and corresponding eigenvector of the square of the Jacobian matrix:

$\begin{pmatrix} \frac{\partial \mathbf{f}}{\partial x}\cdot \frac{\partial \mathbf{f}}{\partial x} & \frac{\partial \mathbf{f}}{\partial x}\cdot \frac{\partial \mathbf{f}}{\partial y} \\ \frac{\partial \mathbf{f}}{\partial x}\cdot \frac{\partial \mathbf{f}}{\partial y} & \frac{\partial \mathbf{f}}{\partial y}\cdot \frac{\partial \mathbf{f}}{\partial y} \\ \end{pmatrix}$

As suggested in the literature, the difference between both eigenvalues should be used as contrast, and the direction of the main eigenvector as gradient direction.

with $\mathbf{f}(x,y)$ being any multi-spectral image. In this case this vector function is defined in a three dimensional color space.

Since this method uses the lti::arctanLUT class to accelerate the computation (sacrifying a little bit in angle precision, which anyway cannot be accurately determined due to the spatial quantization effects) the angles returned in the parameters::Polar mode are always possitive between 0 and 2*Pi.

Constructor & Destructor Documentation

lti::colorContrastGradient::colorContrastGradient ( )

default constructor

lti::colorContrastGradient::colorContrastGradient	(	const parameters::eOutputFormat &	format,
		const int	gradKernelSize = `0`
	)

Construct a functor using gradient kernels with the specified values.

Parameters:

	format	specify the format to be use: Cartesic or Polar
	gradKernelSize	size for the gradient kernel to be used.

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

Construct a functor using the given parameters.

lti::colorContrastGradient::colorContrastGradient ( const colorContrastGradient & other )

copy constructor

Parameters:

other

the object to be copied

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

destructor

Member Function Documentation

bool lti::colorContrastGradient::apply	(	const channel &	c1,
		const channel &	c2,
		const channel &	c3,
		channel &	lambdaMin,
		channel &	lambdaMax,
		channel &	orientation,
		float &	maxMag
	)			const

Computes the eigenvalues of the square of the Jacobian matrix of a tri-spectral image.

This method ignores the parameter format, as it makes only sense for a polar system, and also the contrastFormat parameter as it explicitely requests the computation of all eigenvalues.

Parameters:

	c1	first channel of the input image.
	c2	second channel of the input image.
	c3	third channel of the input image.
	lambdaMin	the resulting minimal eigenvalues
	lambdaMax	the resulting maximal eigenvalues
	orientation	the angle of the main axis (diretion of the eigenvector with the greater eigenvalue)
	maxMag	maximum contrast value found (lambdaMin-lambdaMax)

Returns:: true if apply successful or false otherwise.

bool lti::colorContrastGradient::apply	(	const channel &	c1,
		const channel &	c2,
		channel &	xOrMag,
		channel &	yOrArg,
		float &	maxMag
	)			const

Computes the color contrast gradient of a bi-spectral image defined by the two given components.

Parameters:

	c1	first channel of the input image.
	c2	second channel of the input image.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.
	maxMag	if polar and cartesic modes (see parameters), the maximum magnitude of the gradient found.

Returns:: true if apply successful or false otherwise.

bool lti::colorContrastGradient::apply	(	const channel &	c1,
		const channel &	c2,
		const channel &	c3,
		channel &	xOrMag,
		channel &	yOrArg,
		float &	maxMag
	)			const

Computes the color contrast gradient of a tri-spectral image defined by the three given components.

Parameters:

	c1	first channel of the input image.
	c2	second channel of the input image.
	c3	third channel of the input image.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.
	maxMag	if polar and cartesic modes (see parameters), the maximum magnitude of the gradient found.

Returns:: true if apply successful or false otherwise.

bool lti::colorContrastGradient::apply	(	const channel &	c1,
		const channel &	c2,
		const channel &	c3,
		channel &	xOrMag,
		channel &	yOrArg
	)			const

Computes the color contrast gradient of a tri-spectra image defined by the three given components.

Parameters:

	c1	first channel of the input image.
	c2	second channel of the input image.
	c3	third channel of the input image.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.

Returns:: true if apply successful or false otherwise.

bool lti::colorContrastGradient::apply	(	const channel8 &	src,
		channel &	xOrMag,
		channel &	yOrArg
	)			const

Computes the (monochromatic) gradient of the given channel.

Parameters:

	src	channel with the source data.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.

Returns:: true if apply successful or false otherwise.

Reimplemented from lti::gradientFunctor.

bool lti::colorContrastGradient::apply	(	const channel &	src,
		channel &	xOrMag,
		channel &	yOrArg
	)			const

Computes the (monochromatic) gradient of the given channel.

Parameters:

	src	channel with the source data.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.

Returns:: true if apply successful or false otherwise.

Reimplemented from lti::gradientFunctor.

bool lti::colorContrastGradient::apply	(	const image &	src,
		channel &	xOrMag,
		channel &	yOrArg
	)			const

Computes the color contrast gradient of the given RGB image.

Parameters:

	src	image with the source data. The image will be splitted in its R,G and B components for the analysis.
	xOrMag	the resulting x component or magnitude of the gradient will be left here.
	yOrArg	the resulting y component or argument (orientation) of the gradient will be left here.

Returns:: true if apply successful or false otherwise.

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

returns a pointer to a clone of this functor.

Reimplemented from lti::gradientFunctor.

bool lti::colorContrastGradient::computeColorGradientCart	(	const channel &	c1,
		const channel &	c2,
		const channel &	c3,
		channel &	xchnl,
		channel &	ychnl,
		float &	maxVal
	)			const `[protected]`