---

fl::Transform Class Reference

#include <convolve.h>

Inheritance diagram for fl::Transform:

List of all members.

Public Methods
	Transform (const Matrix< double > &A, bool inverse=false)
	Prefer double format for better inversion (when needed).
	Transform (const Matrix< double > &A, const double scale)
	Assumes A is inverse. Divides first two columns by scale before using A. Just a convenience method.
	Transform (double angle)
	Transform (double scaleX, double scaleY)
void	initialize (const Matrix< double > &A, bool inverse=false)
	A should be at least 2x2.
virtual Image	filter (const Image &image)
	This could be const, but it is useful to allow filters to collect statistics. Note that such filters are not thread safe.
void	setPeg (float centerX=-1, float centerY=-1, int width=-1, int height=-1)
	Set up viewport (of resulting image) so its center hits at a specified point in source image. centerX == -1 means use center of original image. centerY == -1 is similar. width == -1 means use width of original image. height == -1 is similar.
void	setWindow (float centerX, float centerY, int width=-1, int height=-1)
	Set up viewport so its center hits a specified point in what would otherwise be the resulting image.
void	prepareResult (const Image &image, Image &result, Matrix3x3< float > &H)
	Subroutine of filter (). Finalizes parameters that control fit between source and destination images.
Transform	operator * (const Transform &that) const
Public Attributes
Matrix2x2< float >	IA
	Inverse of A. Maps coordinates from output image back to input image.
float	translateX
float	translateY
bool	peg
	Indicates that (centerX, centerY) refers to source image rather than resulting image.
bool	defaultViewport
	Indicates that viewport parameters are calculated rather than provided by user.
float	centerX
float	centerY
int	width
int	height

---

Constructor & Destructor Documentation

Transform::Transform (  const Matrix< double > &    A, bool    inverse = false )

Prefer double format for better inversion (when needed).

Transform::Transform (  const Matrix< double > &    A, const double    scale )

Assumes A is inverse. Divides first two columns by scale before using A. Just a convenience method.

Transform::Transform (  double    angle )

Transform::Transform (  double    scaleX, double    scaleY )

---

Member Function Documentation

Image Transform::filter (  const Image &    image )  [virtual]

This could be const, but it is useful to allow filters to collect statistics. Note that such filters are not thread safe. Implements fl::Filter. Reimplemented in fl::TransformGauss.

void Transform::initialize (  const Matrix< double > &    A, bool    inverse = false )

A should be at least 2x2. Todo: Store full homography rather than separating into translation and deformation components. The only thing preventing this before was some programs depending on a hack for scaling. No program does this any more.

Transform Transform::operator * (  const Transform &    that )  const

void Transform::prepareResult (  const Image &    image, Image &    result, Matrix3x3< float > &    H )

Subroutine of filter (). Finalizes parameters that control fit between source and destination images.

void Transform::setPeg (  float    centerX = -1, float    centerY = -1, int    width = -1, int    height = -1 )

Set up viewport (of resulting image) so its center hits at a specified point in source image. centerX == -1 means use center of original image. centerY == -1 is similar. width == -1 means use width of original image. height == -1 is similar.

void Transform::setWindow (  float    centerX, float    centerY, int    width = -1, int    height = -1 )

Set up viewport so its center hits a specified point in what would otherwise be the resulting image.

---

Member Data Documentation

float fl::Transform::centerX

float fl::Transform::centerY

bool fl::Transform::defaultViewport

Indicates that viewport parameters are calculated rather than provided by user.

int fl::Transform::height

Matrix2x2<float> fl::Transform::IA

Inverse of A. Maps coordinates from output image back to input image.

bool fl::Transform::peg

Indicates that (centerX, centerY) refers to source image rather than resulting image.

float fl::Transform::translateX

float fl::Transform::translateY

int fl::Transform::width

---

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

• convolve.h
• Transform.cc

---