core/vil/algo/vil_corners.h File Reference

---

Detailed Description

Estimate corner positions using Forstner/Harris approach.

Author:

Tim Cootes

Definition in file vil_corners.h.

#include <vil/vil_image_view.h>
#include <vil/algo/vil_sobel_3x3.h>

Go to the source code of this file.

Functions
void	vil_corners (const vil_image_view< float > &grad_i, const vil_image_view< float > &grad_j, vil_image_view< float > &dest, double k=0.04)
	Compute Harris corner strength function given gradient images.
void	vil_corners (const vil_image_view< double > &grad_i, const vil_image_view< double > &grad_j, vil_image_view< double > &dest, double k=0.04)
void	vil_corners_rohr (const vil_image_view< float > &grad_i, const vil_image_view< float > &grad_j, vil_image_view< float > &dest)
	Compute corner strength using Rohr's recommended method.
template<class T>
void	vil_corners (const vil_image_view< T > &src, vil_image_view< float > &dest, double k=0.04)
	Compute Harris corner strength function.
template<class T>
void	vil_corners_rohr (const vil_image_view< T > &src, vil_image_view< float > &dest)
	Compute corner strength using Karl Rohr's recommended method.

---

Function Documentation

template<class T>

void vil_corners	(	const vil_image_view< T > &	src,
		vil_image_view< float > &	dest,
		double	k = 0.04
	)			[inline]

Compute Harris corner strength function.

At each pixel compute the Harris corner function: det(H)-k*sqr(trace(H)), where H is the 2x2 matrix of second derivatives, generated by applying a Sobel operator twice. The filters thus effectively have 5x5 support.

The local peaks of the output image correspond to corner candidates.

Definition at line 49 of file vil_corners.h.

void vil_corners	(	const vil_image_view< double > &	grad_i,
		const vil_image_view< double > &	grad_j,
		vil_image_view< double > &	dest,
		double	k = 0.04
	)

Definition at line 98 of file vil_corners.cxx.

void vil_corners	(	const vil_image_view< float > &	grad_i,
		const vil_image_view< float > &	grad_j,
		vil_image_view< float > &	dest,
		double	k
	)

Compute Harris corner strength function given gradient images.

grad_i and grad_j are assumed to be the i and j gradient images (single plane), such as produced by vil_sobel_3x3(). At each pixel compute the Forstner/Harris corner function: det(H)-k*sqr(trace(H)), where H is the 2x2 matrix of second derivatives, generated by applying a Sobel operator to the gradient images.

The local peaks of the output image correspond to corner candidates.

Definition at line 21 of file vil_corners.cxx.

template<class T>

void vil_corners_rohr	(	const vil_image_view< T > &	src,
		vil_image_view< float > &	dest
	)			[inline]

Compute corner strength using Karl Rohr's recommended method.

This computes the determinant of the matrix C=g.g' after the elements of C have been smoothed. g is the vector of first derivatives (gx,gy)' It relies only on first derivatives.

Definition at line 65 of file vil_corners.h.

void vil_corners_rohr	(	const vil_image_view< float > &	gx,
		const vil_image_view< float > &	gy,
		vil_image_view< float > &	corner_im
	)

Compute corner strength using Rohr's recommended method.

This computes the determinant of the matrix C=g.g' after the elements of C have been smoothed. g is the vector of first derivatives (gx,gy)' It relies only on first derivatives.

Currently uses somewhat inefficient multi-pass method. Could be improved.

Definition at line 183 of file vil_corners.cxx.

---