YAPE feature point detector. More...

#include <yape.h>

Inheritance diagram for yape:

Collaboration diagram for yape:

Classes
struct	dir_table

Public Member Functions
	yape (int width, int height)
	Constructor for detection in `width` x `height` images.
virtual	~yape ()
void	set_radius (int radius)
int	get_radius (void)
void	set_tau (int tau)
int	get_tau (void)
void	activate_bins (void)
void	disactivate_bins (void)
void	set_use_bins (bool p_use_bins)
bool	get_use_bins (void)
void	set_bins_number (int nb_u, int nb_v)
void	activate_subpixel (void)
	Subpixel. Can be activated or disactived (default) for monoscale detection. Always activated for multi-scale detection.
void	disactivate_subpixel (void)
void	set_use_subpixel (bool p_use_subpixel)
void	set_minimal_neighbor_number (int p_minimal_neighbor_number)
int	get_minimal_neighbor_number (void)
int	detect (IplImage image, keypoint points, int max_point_number, IplImage *smoothed_image=0)
void	raw_detect (IplImage *im)
	detect interest points and add them to tmp_points.
int	pick_best_points (keypoint *points, unsigned int max_point_number)
	sort and select the max_point_number best features.
void	save_image_of_detected_points (char name, IplImage image, keypoint *points, int points_nb)
	Save a color image with the detected points on the original image. Useful for visualisation...
IplImage *	get_scores_image (void)
	Return scores and filtered images, for debugging purposes.
IplImage *	get_filtered_image (void)
void	subpix_refine (IplImage im, keypoint p)

Static Public Member Functions
static int	static_detect (IplImage image, keypoint points, int max_point_number, int radius=7, int tau=10)
	Static function for point detection. Slower but avoids an instantiation.

Protected Types
typedef std::vector< keypoint >	keypoint_vector

Protected Member Functions
void	reserve_tmp_arrays (void)
int	get_local_maxima (IplImage *image, int R, float scale)
void	perform_one_point (const unsigned char I, const int x, short Scores, const int Im, const int Ip, const short *dirs, const unsigned char opposite, const unsigned char dirs_nb)
bool	double_check (IplImage image, int x, int y, short dirs, unsigned char dirs_nb)
bool	third_check (const short *Sb, const int next_line)
void	precompute_directions (IplImage image, short _Dirs, int *_Dirs_nb, int R)
void	init_for_monoscale (void)

Protected Attributes
int	minimal_neighbor_number
int	width
int	height
int	radius
int	tau
dir_table *	Dirs
int *	Dirs_nb
keypoint_vector	tmp_points
bool	use_bins
keypoint_vector	bins [10][10]
int	bin_nb_u
int	bin_nb_v
bool	use_subpixel
IplImage *	scores
IplImage *	filtered_image
int	stats_state [100]
int	stats_iter [100]
int	score
int	a
int	b
int	A
int	B0
int	B1
int	B2
int	state
CvMat *	H
CvMat *	mg
CvMat *	shift

Detailed Description

YAPE feature point detector.

Stands for Yet Another Point Extractor.

Usage:

You need OpenCV to use this code. Two different uses of the Yape class are possible:

1) The simplest way is to call the static function static_detect:

int N = yape::static_detect(image, points, maxNumberOfPoints[, radius]);

where:

image is a pointer on a IplImage structure from OpenCv. The type of the image should be IPL_DEPTH_8U, with one channel; The lowest its value is, the finest are the detected points;
points is a array of keypoint s. It should be large enough to store all the detected points;
maxNumberOfPoints is the desired maximum number of detected points.
radius is a parameter for the point detection. Its value should be between 3 and 7, its default value is 7.

The function returns the actual number of detected points.

2) Detection in a large number of images of same size:

If the detection should be performed on a large number of images of same size, you should first create an instance of the detector. The yape constructor performs some pre-computation and pre- memory allocation to save some computation time during the actual detection.

// Create an instance of Yape for detection in 640x480 images:
yape * pe = new yape(640, 480);
// The default value for the radius is 7. Use set_radius() to change it:
pe->set_radius(3);
// Call the function detect() exactly like the static_detect() function:
N[0] = pe->detect(images[0], points[0], 100);
N[1] = pe->detect(images[1], points[1], 100);
N[2] = pe->detect(images[2], points[2], 100);
N[3] = pe->detect(images[3], points[3], 100);
...
// Call the destructor:
delete pe;

The saveImageOfDetectedPoints function is useful to save an image of the detected points:

static void saveImageOfDetectedPoints(char * name, IplImage * image, CvPoint * points, int points_nb);

Definition at line 88 of file yape.h.

Member Typedef Documentation

typedef std::vector<keypoint> yape::keypoint_vector

protected

Definition at line 168 of file yape.h.

Constructor & Destructor Documentation

yape::yape	(	int	width,
		int	height
	)

Constructor for detection in width x height images.

Definition at line 47 of file yape.cpp.

yape::~yape ( )

virtual

Definition at line 66 of file yape.cpp.

Member Function Documentation

void yape::activate_bins ( void )

inline

Definition at line 101 of file yape.h.

References set_use_bins().

void yape::activate_subpixel ( void )

inline

Subpixel. Can be activated or disactived (default) for monoscale detection. Always activated for multi-scale detection.

Definition at line 108 of file yape.h.

References set_use_subpixel().

int yape::detect	(	IplImage *	image,
		keypoint *	points,
		int	max_point_number,
		IplImage *	smoothed_image = `0`
	)

Definition at line 478 of file yape.cpp.

Referenced by static_detect().

void yape::disactivate_bins ( void )

inline

Definition at line 102 of file yape.h.

References set_use_bins().

void yape::disactivate_subpixel ( void )

inline

Definition at line 109 of file yape.h.

References set_use_subpixel().

bool yape::double_check	(	IplImage *	image,
		int	x,
		int	y,
		short *	dirs,
		unsigned char	dirs_nb
	)

protected

Definition at line 196 of file yape.cpp.

IplImage* yape::get_filtered_image ( void )

inline

Definition at line 131 of file yape.h.

References filtered_image.

int yape::get_local_maxima	(	IplImage *	image,
		int	R,
		float	scale
	)

protected

Find local maximas in score image and append them to tmp_points.

Returns: the number of local maxima found.

Definition at line 684 of file yape.cpp.

References is_local_maxima(), keypoint::scale, keypoint::score, keypoint::u, keypoint::v, and yape_bin_size.

Referenced by pyr_yape::detect().

int yape::get_minimal_neighbor_number ( void )

inline

Definition at line 113 of file yape.h.

References minimal_neighbor_number.

int yape::get_radius ( void )

inline

Definition at line 96 of file yape.h.

References radius.

IplImage* yape::get_scores_image ( void )

inline

Return scores and filtered images, for debugging purposes.

Definition at line 130 of file yape.h.

References scores.

int yape::get_tau ( void )

inline

Definition at line 99 of file yape.h.

References tau.

bool yape::get_use_bins ( void )

inline

Definition at line 104 of file yape.h.

References use_bins.

void yape::init_for_monoscale ( void )

protected

Definition at line 97 of file yape.cpp.

References yape_max_radius.

void yape::perform_one_point	(	const unsigned char *	I,
		const int	x,
		short *	Scores,
		const int	Im,
		const int	Ip,
		const short *	dirs,
		const unsigned char	opposite,
		const unsigned char	dirs_nb
	)

protected

Definition at line 253 of file yape.cpp.

References A_INF, A_NOT_INF, A_NOT_SUP, A_SUP, B0_INF, B0_NOT_INF, B0_NOT_SUP, B0_SUP, B1_EQUAL_B2_NOT_INF, B1_EQUAL_B2_NOT_SUP, B1_INF, B1_INF_B2_INF, B1_INF_B2_NOT_SUP, B1_INF_B2_SUP, B1_NOT_INF, B1_NOT_INF_B2_NOT_INF, B1_NOT_SUP, B1_NOT_SUP_B2_NOT_SUP, B1_SUP, B1_SUP_B2_INF, B1_SUP_B2_NOT_INF, B1_SUP_B2_SUP, B2_INF, B2_NOT_INF, B2_NOT_SUP, B2_SUP, GET_A, GET_B0, GET_B1, GET_B2, GOTO_END_NOT_AN_INTEREST_POINT, GOTO_STATE, and PUT_B2_IN_B1_AND_GET_B2.

int yape::pick_best_points	(	keypoint *	points,
		unsigned int	max_point_number
	)

sort and select the max_point_number best features.

This method sorts tmp_points and select the max_point_number best points, as long as the score is high enough.

Definition at line 518 of file yape.cpp.

References mymin(), and N.

Referenced by pyr_yape::detect().

void yape::precompute_directions	(	IplImage *	image,
		short *	_Dirs,
		int *	_Dirs_nb,
		int	R
	)

protected

Definition at line 126 of file yape.cpp.

Referenced by pyr_yape::pyr_yape().

void yape::raw_detect ( IplImage * im )

detect interest points and add them to tmp_points.

Detect interest points, without filtering and without selecting best ones. Just find them and add them to tmp_points. tmp_points is not cleared in this method.

Definition at line 573 of file yape.cpp.

Referenced by pyr_yape::detect().

void yape::reserve_tmp_arrays ( void )

protected

Definition at line 110 of file yape.cpp.

References yape_bin_size, and yape_tmp_points_array_size.

Referenced by pyr_yape::detect().