---

fl::FilterHarris Class Reference

#include <convolve.h>

Inheritance diagram for fl::FilterHarris:

List of all members.

Public Methods
	FilterHarris (double sigmaD=1.0, double sigmaI=1.4, const PixelFormat &format=GrayFloat)
virtual Image	filter (const Image &image)
	Relies on preprocess, process, and response to do all the work.
virtual void	preprocess (const Image &image)
	Extracts the square gradient matrices from the image. Stores in xx, xy, and yy.
virtual Image	process ()
	Collects responses into an Image.
virtual double	response (int x, int y) const
	Returns Harris function value (of mose recently filtered image) at (x,y). The location uses the same coordinates as the result of filter(). Uses squareGradient().
virtual void	gradientSquared (int x, int y, Matrix< double > &result) const
	Finds the autocorrelation matrix (of the most recently filtered image) at (x,y). The location uses the same coordinates as the result of filter().
Public Attributes
double	sigmaD
	Derivation scale.
double	sigmaI
	Integration scale.
Gaussian2D	G_I
	Gaussian for integration.
Gaussian1D	G1_I
	seperated Gaussian for integration
Gaussian1D	G1_D
	seperated Gaussian for derivation (blurring pass)
GaussianDerivative1D	dG_D
	seperated Gaussian for derivation
Image	xx
	Components of the autocorrelation matrix.
Image	xy
	These are built by preprocess ().
Image	yy
int	offset
	Total amount of one image border removed.
int	offsetI
	Border removed by integration.
int	offsetD
	Border removed by differentiation.
Static Public Attributes
const double	alpha = 0.06
Protected Attributes
int	offset1
int	offset2

---

Constructor & Destructor Documentation

FilterHarris::FilterHarris (  double    sigmaD = 1.0, double    sigmaI = 1.4, const PixelFormat &    format = GrayFloat )

---

Member Function Documentation

Image FilterHarris::filter (  const Image &    image )  [virtual]

Relies on preprocess, process, and response to do all the work. Implements fl::Filter.

void FilterHarris::gradientSquared (  int    x, int    y, Matrix< double > &    result )  const [virtual]

Finds the autocorrelation matrix (of the most recently filtered image) at (x,y). The location uses the same coordinates as the result of filter().

void FilterHarris::preprocess (  const Image &    image )  [virtual]

Extracts the square gradient matrices from the image. Stores in xx, xy, and yy.

Image FilterHarris::process (    )  [virtual]

Collects responses into an Image. Reimplemented in fl::FilterHarrisEigen.

double FilterHarris::response (  int    x, int    y )  const [virtual]

Returns Harris function value (of mose recently filtered image) at (x,y). The location uses the same coordinates as the result of filter(). Uses squareGradient(). Reimplemented in fl::FilterHarrisEigen.

---

Member Data Documentation

const double FilterHarris::alpha = 0.06 [static]

GaussianDerivative1D fl::FilterHarris::dG_D

seperated Gaussian for derivation

Gaussian1D fl::FilterHarris::G1_D

seperated Gaussian for derivation (blurring pass)

Gaussian1D fl::FilterHarris::G1_I

seperated Gaussian for integration

Gaussian2D fl::FilterHarris::G_I

Gaussian for integration.

int fl::FilterHarris::offset

Total amount of one image border removed.

int fl::FilterHarris::offset1 [protected]

If the blurring part of the separable Gaussian derivative kernel has a larger radius, then the difference in pixels is stored in offset1. If the derivative part has a larger radius, then the difference is kept in offset2. These help align the x and y derivative images correctly.

int fl::FilterHarris::offset2 [protected]

int fl::FilterHarris::offsetD

Border removed by differentiation.

int fl::FilterHarris::offsetI

Border removed by integration.

double fl::FilterHarris::sigmaD

Derivation scale.

double fl::FilterHarris::sigmaI

Integration scale.

Image fl::FilterHarris::xx

Components of the autocorrelation matrix.

Image fl::FilterHarris::xy

These are built by preprocess ().

Image fl::FilterHarris::yy

---

The documentation for this class was generated from the following files:

• convolve.h
• FilterHarris.cc

---