OkanReg Class Reference

#include <OkanReg.h>

List of all members.

Public Member Functions
double	norCrosCorr (int x0, int y0, int x1, int y1, int n, double mean0, double std0, double mean1, double std1)
void	findHomography (int *ids, CvMat *h)
int	checkHomographyQuality (CvMat *h, float errTol, float &mse)
	OkanReg ()
	~OkanReg ()
void	clearReg ()
void	destroyImage (int id)
void	setImage (IplImage *im, int id)
void	setImage (string im, int id)
IplImage *	getImage (int id)
void	setImageROI (int id, int x1, int y1, int x2, int y2)
CvRect	getImageROI (int id)
void	removeImageROI (int id)
int	getCornerSize (int id)
CvPoint2D32f *	getCorner (int id)
void	clearMatches ()
void	addMatch (CvPoint2D32f p1, CvPoint2D32f p2)
void	eraseMatch (int id)
vector< CvPoint2D32f > &	getMatches (int id)
CvMat *	getBestMat (int id)
int	getBestPoint ()
float	getBestMse ()
double	findHarris (int id, int n, double thresh, double minDistance, int blockSize=3, int apertureSize=3, double k=0.04)
int	findPutMatch (int n, int d=0, double thresh=-2)
bool	findRANSACHomography (int iter, float errTol)
Public Attributes
IplImage *	image [2]
IplImage *	gray [2]
CvPoint2D32f *	corner [2]
int	cornerSize [2]
vector< CvPoint2D32f >	matches [2]
CvMat *	bestMat [2]
int	bestPoint
float	bestMse

---

Detailed Description

Image Registration class

Author:

Yunus OLGUN

---

Constructor & Destructor Documentation

OkanReg::OkanReg

(

)

Constructor of OkanReg class. It simply makes all values 0 and all pointers NULL.

OkanReg::~OkanReg

(

)

Destructor of OkanReg class. It releases all data it has. Be carefull about that, it does not releases image field of the class.

See also:

destroyImage()

---

Member Function Documentation

double OkanReg::norCrosCorr	(	int	x0,
		int	y0,
		int	x1,
		int	y1,
		int	n,
		double	mean0,
		double	std0,
		double	mean1,
		double	std1
	)

Private member function to calculate normalized cross correlation of two points.

Parameters:

	x0	x coordinate of first point
	y0	y coordinate of first point
	x1	x coordinate of second point
	y1	y coordinate of second point
	n	window size for calculating mean and standard deviation of a corner
	mean0	mean of first point
	std0	standard deviation of first point
	mean1	mean of second point
	std1	standard deviation of second point

Returns:

normalized cross correlation of two points

void OkanReg::findHomography	(	int *	ids,
		CvMat *	h
	)

Finds homography matrix for one subset of putative matches.

Parameters:

	ids	It contains 4 integer which are indexes of putative matches
	h	A CvMat pointer to be filled by this function. It is used as return value.

int OkanReg::checkHomographyQuality	(	CvMat *	h,
		float	errTol,
		float &	mse
	)

A private function to calculate quality of a homography matrix

Parameters:

	h	the homography matrix
	errTol	error tolerance while calculating quality of a homography matrix
	mse	mean square error of homography matrix. It is used as a return value

Returns:

quality of homography matrix

void OkanReg::clearReg

(

)

It does same work of a destructor.

See also:

~OkanReg()

clearReg()

void OkanReg::destroyImage

(

int

id

)

It releases all data of the image which has the id. It is used for destructor and clearReg also.

Parameters:

id

id of image

See also:

~OkanReg()

~clearReg()

void OkanReg::setImage	(	IplImage *	im,
		int	id
	)

It sets im as a new image for registration process. It does not copy all image. It only copy the pointer. It releases all data for previous image named id. It calls destroyImage() for this process.

Parameters:

	im	pointer of new image
	id	the of image to be setted. It should 0 or 1

See also:

destroyImage()

void OkanReg::setImage	(	string	im,
		int	id
	)

Todo:

DO NOT USE THIS FUNCTION. While removing it do not forget to remove highgui.h from includes!!!

IplImage * OkanReg::getImage

(

int

id

)

Get method of image field

Parameters:

id

the id of the image which will be returned

Returns:

pointer to the this image

void OkanReg::setImageROI	(	int	id,
		int	x1,
		int	y1,
		int	x2,
		int	y2
	)

sets region of interest for the image. Give a coordination of rectangle. Be carefull about that it affects every computation.

Parameters:

	id	image id
	x1	x coordinate of top-left of rectangle
	y1	y coordinate of top-left of rectangle
	x2	width of rectangle
	y2	heigth of rectangle

CvRect OkanReg::getImageROI

(

int

id

)

Get method for region of interest of image

Parameters:

id

image id

Returns:

CvRect which contains rectangle for region of interest

void OkanReg::removeImageROI

(

int

id

)

Removes region of interest from image.

Parameters:

id

image id

int OkanReg::getCornerSize

(

int

id

)

Returns size of corners

Parameters:

id

image id

Returns:

size of corners

CvPoint2D32f * OkanReg::getCorner

(

int

id

)

Get method of corner field.

Parameters:

id

image id

Returns:

return an array which contains corners.

void OkanReg::clearMatches

(

)

Clears putative matches of two images. Because they are related with each other two of them will be removed at once.

void OkanReg::addMatch	(	CvPoint2D32f	p1,
		CvPoint2D32f	p2
	)

Adds a putative match to matches field.

Parameters:

	p1	point for first image
	p2	point for second image

void OkanReg::eraseMatch

(

int

id

)

Removes the putative match from matches field.

Parameters:

id

index of the putative match

vector< CvPoint2D32f > & OkanReg::getMatches

(

int

id

)

Get method of images

Parameters:

id

image id

Returns:

vector of putative matches

CvMat * OkanReg::getBestMat

(

int

id

)

Get method of best homography matrix

Parameters:

id

image id

Returns:

best homography matrix

int OkanReg::getBestPoint

(

)

Returns the number of putative matches which correspond to best homography matrix

Returns:

the number of putative matches which correspond to best homography matrix

float OkanReg::getBestMse

(

)

Returns mean square error of best homography matrix

Returns:

mean square error of best homography matrix

double OkanReg::findHarris	(	int	id,
		int	n,
		double	thresh,
		double	minDistance,
		int	blockSize = 3,
		int	apertureSize = 3,
		double	k = 0.04
	)

Finds corners of the given image and pushes them into the corner field and sets cornerSize. ROI is effective at this.

Parameters:

	id	image id
	n	how many points should be found. If n is equal to 0, then thresh is ignored.
	thresh	quality level for determining weak corener points.
	minDistance	minimum distance between two corner points. According to this weaker ones will be eliminated from result.
	blockSize	block size of image. Default value is 3.
	apertureSize	aperture size for process. Default value is 3.
	k	a parameter for the corner detection algorithm. Default value is 0.04

Returns:

selected weakest corner's quality level

int OkanReg::findPutMatch	(	int	n,
		int	d = 0,
		double	thresh = -2
	)

Finds putative matches of two images and pushes them into the matches field. If cornerSize is 0 it throws an exception. ROI is not effective

Parameters:

	n	window size for calculating mean and standard deviation of a corner.
	d	maximum distance between two corners to be a putative match.
	thresh	thresh value for eliminating weak putative matches.

Returns:

size of found putative matches

bool OkanReg::findRANSACHomography	(	int	iter,
		float	errTol
	)

Finds best homography matrix according to RANSAC algorithm. If size of 4-combinations is less than iter it tries every possible one. Otherwise it applies RANSAC algorithm

Parameters:

	iter	maximum number of iterations for RANSAC algorithm
	errTol	error tolerance for homography matrix. It is used in determining quality of homography matrix. If the difference between matches is smaller than this errTol variable, homography matrix's quality increases 1 point

Returns:

If successfull it return true, otherwise false

---

Member Data Documentation

IplImage* OkanReg::image[2]

Private IPLImage* array which keeps registered images.

IplImage* OkanReg::gray[2]

Private IPLImage* array which keeps gray level of registered images.

CvPoint2D32f* OkanReg::corner[2]

Private CVPoint2D32f* array which keeps corner points of two images.

int OkanReg::cornerSize[2]

Private int array which keeps size of corner points of two images.

vector< CvPoint2D32f > OkanReg::matches[2]

Private CvPoint2D32f vector which is used for putative matches of two images.

CvMat* OkanReg::bestMat[2]

Private CVMat* array which keeps best homography matrix of two images. bestMat[0] means image[0] * bestMat[0] => image[1]

int OkanReg::bestPoint

Private int which is used for estimating value of bestMat homography matrix. It means that how many of putative matches corresponds to best homography matrix.

float OkanReg::bestMse

Private float which is used for mean square error of best homography matrix

---

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

• include/OkanReg.h
• src/OkanReg.cpp

---