Conversion between ImFusion and OpenCV image containers and primitives. More...

Detailed Description

Conversion between ImFusion and OpenCV image containers and primitives.

Functions
bool	convert (MemImage *imgIn, cv::Mat &matOut, bool container=true)
	Convert 2D image to OpenCV, possibly in container mode Warning: Container mode does not support correctly colored images (RGB vs BGR).
cv::Mat	convert (MemImage *img, bool container=true)
std::unique_ptr< TypedImage< double > >	convert (cv::Mat &matIn)
std::unique_ptr< MemImage >	convert (cv::Mat &matIn, std::optional< PixelType > targetType, bool convertBGR2RGB=true)
	Converts an OpenCV image to a MemImage.
std::optional< mat4 >	convert (const cv::Mat &rotation, const cv::Mat &translation)
	Converts an OpenCV pose representation (rotation and translation) into a homogeneous SE(3) transform (4x4).
bool	convert (const mat4 &se3Transform, cv::Mat &rvecOut, cv::Mat &tvecOut)
	Converts a homogeneous SE(3) transform (4x4) into OpenCV pose parameters (Rodrigues rvec and translation tvec).
int	type (const Image *desc)
	Get OpenCV data type from ImFusion type and channels.
std::vector< cv::Point2f >	convertToPoint2fVector (const std::vector< vec2f > &p)
std::vector< cv::Point2d >	convertToPoint2dVector (const std::vector< vec2f > &p)
std::vector< cv::Point2f >	convertToPoint2fVector (const std::vector< vec2 > &p)
std::vector< cv::Point2d >	convertToPoint2dVector (const std::vector< vec2 > &p)
std::vector< cv::Point3f >	convertToPoint3fVector (const std::vector< vec3f > &p)
std::vector< cv::Point3d >	convertToPoint3dVector (const std::vector< vec3f > &p)
std::vector< cv::Point3f >	convertToPoint3fVector (const std::vector< vec3 > &p)
std::vector< cv::Point3d >	convertToPoint3dVector (const std::vector< vec3 > &p)
std::vector< vec2 >	convertFromPoint2fVector (const std::vector< cv::Point2f > &p)
std::vector< vec2 >	convertFromPoint2dVector (const std::vector< cv::Point2d > &p)
std::vector< vec3 >	convertFromPoint3fVector (const std::vector< cv::Point3f > &p)
std::vector< vec3 >	convertFromPoint3dVector (const std::vector< cv::Point3d > &p)
std::shared_ptr< GlDenseDeformation >	convertOpenCVDistortionToDeformation (const Image &imgDesc, const std::array< double, 5 > &distCoeffs, const mat3 &Kin, const mat3 &Kout, bool rescaleByExtent=true, bool flipY=false)
	Converts OpenCV undistort map to GlDenseDeformation.
std::shared_ptr< GlDenseDeformation >	convertOpenCVDistortionToInverseDeformation (int width, int height, const std::array< double, 5 > &distCoeffs, const mat3 &Kin, bool flipY=false)
	Obtains a deformation field to model camera distortion using OpenCV.

Function Documentation

◆ convert() [1/3]

std::unique_ptr< MemImage > ImFusion::OpenCV::convert	(	cv::Mat &	matIn,
		std::optional< PixelType >	targetType,
		bool	convertBGR2RGB = true )

Converts an OpenCV image to a MemImage.

Parameters

matIn	The input image
targetType	The PixelType the returned image should have, if not given a target format is chosen that matches the pixel format of matIn. (default is std::nullopt)
convertBGR2RGB	Only for three-channel color images: whether to convert the OpenCV-default BGR format to the RGB format used in MemImage (default is true)

Returns: a new MemImage

◆ convert() [2/3]

std::optional< mat4 > ImFusion::OpenCV::convert	(	const cv::Mat &	rotation,
		const cv::Mat &	translation )

Converts an OpenCV pose representation (rotation and translation) into a homogeneous SE(3) transform (4x4).

This function accepts a rotation either as a 3x3 rotation matrix \(R\) or as a Rodrigues rotation vector \(\omega\) (3x1 or 1x3). The translation must be a 3-element vector (3x1 or 1x3).

Note: Input matrices must be single-channel and of floating-point type (CV_32F or CV_64F). Conversion between float and double is handled internally by OpenCV/Eigen conversion utilities, but non-floating types are rejected.; If rotation is provided as a Rodrigues vector, it is converted to a 3x3 rotation matrix via cv::Rodrigues().; If translation is provided as a row vector (1x3), it is transposed internally to match a column vector.

Parameters

rotation	Rotation as either: 3x3 matrix or 3-element vector (shape 3x1 or 1x3). Must be single-channel CV_32F or CV_64F.
translation	Translation vector (shape 3x1 or 1x3). Must be single-channel CV_32F or CV_64F.

Returns: A 4x4 homogeneous transform on success; std::nullopt on invalid input (wrong type/shape/empty).

◆ convert() [3/3]

bool ImFusion::OpenCV::convert	(	const mat4 &	se3Transform,
		cv::Mat &	rvecOut,
		cv::Mat &	tvecOut )

Converts a homogeneous SE(3) transform (4x4) into OpenCV pose parameters (Rodrigues rvec and translation tvec).

Note

The routine validates that the rotation block is (approximately) a proper rotation matrix by checking:

\(\det(R) \approx 1\)
\(R R^T \approx I\)

Parameters

se3Transform	Homogeneous 4x4 transform. The top-left 3x3 block is interpreted as rotation, and the top-right 3x1 block as translation.
rvecOut	Output Rodrigues rotation vector, a 3x1 CV_64F matrix
tvecOut	Output translation vector, a 3x1 CV_64F matrix.

Returns: true on success, false if the rotation is not orthonormal and positive-definite.

◆ convertOpenCVDistortionToDeformation()

std::shared_ptr< GlDenseDeformation > ImFusion::OpenCV::convertOpenCVDistortionToDeformation	(	const Image &	imgDesc,
		const std::array< double, 5 > &	distCoeffs,
		const mat3 &	Kin,
		const mat3 &	Kout,
		bool	rescaleByExtent = true,
		bool	flipY = false )

Converts OpenCV undistort map to GlDenseDeformation.

See also: cv::initUndistortRectifyMap If

Parameters

rescaleByExtent	is false, output deformation is in pixel coordinates. Otherwise output deformation is in pixel coordinates component wise divided by extent of
imgDesc.
flipY	determins whether to flip the Y coordinate from OpenCV representation

◆ convertOpenCVDistortionToInverseDeformation()

std::shared_ptr< GlDenseDeformation > ImFusion::OpenCV::convertOpenCVDistortionToInverseDeformation	(	int	width,
		int	height,
		const std::array< double, 5 > &	distCoeffs,
		const mat3 &	Kin,
		bool	flipY = false )

Obtains a deformation field to model camera distortion using OpenCV.

Parameters

distCoeffs	are OpenCV-style distortion coefficients
Kin	is the camera matrix (onto pixel coordinates).
flipY	determins whether to flip the Y coordinate from OpenCV representation

See also: cv::initUndistortRectifyMap Warning: output deformation is in pixel coordinates.

Detailed Description

Functions

Function Documentation

◆ convert() [1/3]

◆ convert() [2/3]

◆ convert() [3/3]

◆ convertOpenCVDistortionToDeformation()

◆ convertOpenCVDistortionToInverseDeformation()