DVision::HSolver Class Reference

Computes fundamental matrices. More...

#include <HSolver.h>

Inheritance diagram for DVision::HSolver:

List of all members.

Public Member Functions
	HSolver ()
	HSolver (int cols, int rows)
virtual	~HSolver ()
virtual void	setImageSize (int cols, int rows)
cv::Mat	findHomography (const cv::Mat &P1, const cv::Mat &P2, double reprojection_error, int min_points=5, std::vector< uchar > *status=NULL, bool computeH=true, double probability=0.99, int max_its=500) const
bool	checkHomography (const cv::Mat &P1, const cv::Mat &P2, double reprojection_error, int min_points=5, double probability=0.99, int max_its=500) const
Protected Member Functions
cv::Mat	_computeH (const cv::Mat &Qc1, const cv::Mat &Qc2, const std::vector< unsigned int > &i_cols) const

---

Detailed Description

Computes fundamental matrices.

---

Constructor & Destructor Documentation

DVision::HSolver::HSolver

(

)

Creates the solver without setting the image dimensions

DVision::HSolver::HSolver	(	int	cols,
		int	rows
	)

Creates the solver and set the image dimensions

Parameters:

	cols	width of images
	rows	height of images

virtual DVision::HSolver::~HSolver

(

)

[inline, virtual]

Destructor

---

Member Function Documentation

cv::Mat DVision::HSolver::_computeH	(	const cv::Mat &	Qc1,
		const cv::Mat &	Qc2,
		const std::vector< unsigned int > &	i_cols
	)			const [protected]

Computes H from correspondences Q1(:,i_cols), Q2(:,i_cols)

Parameters:

	Qc1	3xN normalized
	Qc2	3xN normalized
	i_cols	# >= 4

Returns:

H12 3x3 or empty

bool DVision::HSolver::checkHomography	(	const cv::Mat &	P1,
		const cv::Mat &	P2,
		double	reprojection_error,
		int	min_points = 5,
		double	probability = 0.99,
		int	max_its = 500
	)			const

Checks if a consistent homography matrix can be computed from the given points. It is not computed, though.

Parameters:

	P1	2xN, 3xN, Nx2, Nx3, correspondences of image 1 in image coordinates
	P2	2xN, 3xN, Nx2, Nx3, correspondences of image 2 in image coordinates
	reprojection_error	max reprojection error for getting inliers
	min_points	min number of required inliers
	probability	RANSAC success probability
	max_its	maximum number of RANSAC iterations

Returns:

true iff some fundamental matrix is found

cv::Mat DVision::HSolver::findHomography	(	const cv::Mat &	P1,
		const cv::Mat &	P2,
		double	reprojection_error,
		int	min_points = 5,
		std::vector< uchar > *	status = NULL,
		bool	computeH = true,
		double	probability = 0.99,
		int	max_its = 500
	)			const

Finds an homography matrix from the given correspondences by running RANSAC

Parameters:

	P1	2xN, 3xN, Nx2, Nx3, correspondences of image 1 in image coordinates
	P2	2xN, 3xN, Nx2, Nx3, correspondences of image 2 in image coordinates
	reprojection_error	max reprojection error for getting inliers
	min_points	min number of required inliers
	status	(out) vector s.t. status[i] == 1 if i-th point is an inlier, 0 otherwise
	computeH	(default: true) if false, the final H is not computed and an arbitrary non-empty 3x3 matrix is returned (it saves a svd operation)
	probability	RANSAC success probability
	max_its	maximum number of RANSAC iterations

Returns:

H s.t. P1' * H * P2 == 0, or empty

void DVision::HSolver::setImageSize	(	int	cols,
		int	rows
	)			[virtual]

Sets image size

Parameters:

	cols
	rows	FSolver::setImageSize

Reimplemented from DVision::FSolver.

---

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

• HSolver.h
• HSolver.cpp