Image Transforms

Functions for transforming images by rotating, translating, and magnifying.

Classes
struct	mx::improc::bilinearTransform< _arithT >
	Transformation by bi-linear interpolation. More...
struct	mx::improc::cubicConvolTransform< _arithT >
	Transformation by cubic convolution interpolation. More...

Typedefs
typedef bilinearTransform< float >	mx::improc::bilinearTransf
	Typedef for bilinearTransform with single precision.
typedef bilinearTransform< double >	mx::improc::bilinearTransd
	Typedef for bilinearTransform with double precision.
typedef cubicConvolTransform< float >	mx::improc::cubicConvolTransf
	Typedef for cubicConvolTransform with single precision.
typedef cubicConvolTransform< double >	mx::improc::cubicConvolTransd
	Typedef for cubicConvolTransform with double precision.

Typedef Documentation

bilinearTransd

typedef bilinearTransform<double> mx::improc::bilinearTransd

Typedef for bilinearTransform with double precision.

Definition at line 72 of file imageTransforms.hpp.

bilinearTransf

typedef bilinearTransform<float> mx::improc::bilinearTransf

Typedef for bilinearTransform with single precision.

Definition at line 67 of file imageTransforms.hpp.

cubicConvolTransd

typedef cubicConvolTransform<double> mx::improc::cubicConvolTransd

Typedef for cubicConvolTransform with double precision.

Definition at line 178 of file imageTransforms.hpp.

cubicConvolTransf

typedef cubicConvolTransform<float> mx::improc::cubicConvolTransf

Typedef for cubicConvolTransform with single precision.

Definition at line 173 of file imageTransforms.hpp.

Function Documentation

imageDownSample()

template<typename imageOutT , typename imageInT >

void mx::improc::imageDownSample	(	imageOutT &	imout,
		const imageInT &	imin
	)

Down-sample an image, reducing its size while conserving the total flux.

If the old size is an integer multiple of the new size, this is just a re-bin. If not an integer multiple, the image is interpolated after performing the closest re-bin, and then re-normalized to conserve flux.

Todo:

Allow selection of interpolator, providing a default version.

Definition at line 874 of file imageTransforms.hpp.

Referenced by mx::AO::sim::ccdDetector< _realT >::exposeImage().

imageMagnify() [1/2]

template<typename arrOutT , typename arrInT >

void mx::improc::imageMagnify	(	arrOutT &	transim,
		const arrInT &	im
	)

Magnify an image with the cubic convolution interpolator.

Uses the cubic convolution interpolator to magnify the input image to the size of the output image.

This is a wrapper for imageMagnify with the transform type specified.

Template Parameters

arrOutT	is the eigen array type of the output.
arrInT	is the eigen array type of the input.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

[out]	transim	contains the magnified image. Must be pre-allocated.
[in]	im	is the image to be magnified.

Definition at line 660 of file imageTransforms.hpp.

References mx::improc::imageMagnify().

imageMagnify() [2/2]

template<typename arrOutT , typename arrInT , typename transformT >

void mx::improc::imageMagnify	(	arrOutT &	transim,
		const arrInT &	im,
		transformT	trans
	)

Magnify an image.

Uses the given transformation type to magnify the input image to the size of the output image.

Here we assume that the image center is the mxlib standard:

x_center = 0.5*(im.rows()-1);
y_center = 0.5*(im.cols()-1);

Some care is necessary to prevent magnification from shifting the image with respect to this center. The main result is that the magnification factors (which can be different in x and y) are defined thus:

x_mag = (transim.rows()-1.0) / (im.rows()-1.0);
y_mag = (transim.cols()-1.0) / (im.cols()-1.0);

Example:

im1.resize(512,512);
//add image to im1
im2.resize(1024,1024);
imageMagnify(im2,im1, cubicConvolTransform<double>());

In this exmple, the image in im1 will be magnified by 1023.0/511.0 = 2.002x and placed in im2.

This transform function does not handle edges. If treatment of edges is desired, you must pad the input image using the desired strategy before calling this function. Note that the padded-size of the input image will affect the magnification factor.

Template Parameters

arrOutT	is the eigen array type of the output.
arrInT	is the eigen array type of the input.
transformT	specifies the transformation to use.

Parameters

[out]	transim	contains the magnified image. Must be pre-allocated.
[in]	im	is the image to be magnified.
[in]	trans	is the transformation to use

Definition at line 571 of file imageTransforms.hpp.

Referenced by mx::improc::imageXCorrFFT< _realImageT >::findPeak(), mx::improc::imageMagnify(), mx::improc::imageMaxInterp(), and mx::improc::imageXCorrDiscrete< _ccImT >::operator()().

imageRebinMean() [1/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinMean	(	imageOutT &	imout,
		const imageInT &	imin
	)

Re-bin an image using the mean.

This is a wrapper for imageRebinSum with mean=true.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[in]	imin	the image to rebin

Definition at line 744 of file imageTransforms.hpp.

References mx::improc::imageRebinSum().

imageRebinMean() [2/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinMean	(	imageOutT &	imout,
		int &	xMax,
		int &	yMax,
		typename imageOutT::Scalar &	pMax,
		const imageInT &	imin,
		bool	mean = false
	)

Re-bin an image using the mean. Records the value and position of the re-binned max pixel.

This is a wrapper for imageRebinSum with mean=true.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[out]	xMax	the x-locatioin of the max pixel
[out]	yMax	the y-locatioin of the max pixel
[out]	pMax	the value of the max pixel
[in]	imin	the image to rebin
[in]	mean	if true the output is the mean rather than the sum.

Definition at line 758 of file imageTransforms.hpp.

References mx::improc::imageRebinSum().

imageRebinMeanReject() [1/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinMeanReject	(	imageOutT &	imout,
		const imageInT &	imin
	)

Re-bin an image, takes the mean with a min/max rejection.

The mean is calculated after rejecting the minimuma and maximum value.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[in]	imin	the image to rebin

Definition at line 774 of file imageTransforms.hpp.

imageRebinMeanReject() [2/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinMeanReject	(	imageOutT &	imout,
		int &	xMax,
		int &	yMax,
		typename imageOutT::Scalar &	pMax,
		const imageInT &	imin
	)

Re-bin an image, takes the mean with a min/max rejection. Records the value and position of the re-binned max pixel. The mean is calculated after rejecting the minimuma and maximum value.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[out]	xMax	the x-locatioin of the max pixel
[out]	yMax	the y-locatioin of the max pixel
[out]	pMax	the value of the max pixel
[in]	imin	the image to rebin

Definition at line 817 of file imageTransforms.hpp.

imageRebinSum() [1/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinSum	(	imageOutT &	imout,
		const imageInT &	imin,
		bool	mean = false
	)

Re-bin an image using the sum, reducing its size while conserving the total flux.

Optionally this can be the mean instead of the sum filter, in which case total flux is not conserved.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[in]	imin	the image to rebin
[in]	mean	if true the output is the mean rather than the sum.

Definition at line 671 of file imageTransforms.hpp.

Referenced by mx::improc::imageRebinMean(), and mx::improc::imageRebinMean().

imageRebinSum() [2/2]

template<typename imageOutT , typename imageInT >

int mx::improc::imageRebinSum	(	imageOutT &	imout,
		int &	xMax,
		int &	yMax,
		typename imageOutT::Scalar &	pMax,
		const imageInT &	imin,
		bool	mean = false
	)

Re-bin an image using the sum, reducing its size while conserving the total flux. Records the value and position of the re-binned max pixel. Optionally this can be the mean instead of the sum filter, in which case total flux is not conserved.

This is an overloaded member function, provided for convenience. It differs from the above function only in what argument(s) it accepts.

Parameters

[out]	imout	the re-binned image. Must be allocated to size which is an integer factor smaller than imin.
[out]	xMax	the x-locatioin of the max pixel
[out]	yMax	the y-locatioin of the max pixel
[out]	pMax	the value of the max pixel
[in]	imin	the image to rebin
[in]	mean	if true the output is the mean rather than the sum.

Definition at line 702 of file imageTransforms.hpp.

imageRotate()

template<typename transformT , typename arrT , typename arrT2 , typename floatT >

void mx::improc::imageRotate	(	arrT &	transim,
		const arrT2 &	im,
		floatT	dq,
		transformT	trans
	)

Rotate an image represented as an eigen array.

Uses the given transformation type to rotate an image.

Template Parameters

transformT	specifies the transformation to use [will be resolved by compiler]
arrT	is the eigen array type of the output [will be resolved by compiler]
arrT2	is the eigen array type of the input [will be resolved by compiler]
floatT	is a floating point type [will be resolved by compiler in most cases]

Parameters

[out]	transim	The rotated image. Must be pre-allocated.
[in]	im	The image to be rotated.
[in]	dq	the angle, in radians, by which to rotate in the c.c.w. direction
[in]	trans	is the transformation to use

Definition at line 194 of file imageTransforms.hpp.

Referenced by mx::improc::ADIobservation< _realT, _derotFunctObj >::derotate(), and mx::improc::rotateMask().

imageShift()

template<typename arrOutT , typename arrInT , typename floatT1 , typename floatT2 , typename transformT >

void mx::improc::imageShift	(	arrOutT &	transim,
		const arrInT &	im,
		floatT1	dx,
		floatT2	dy,
		transformT	trans
	)

Shift an image.

Uses the given transformation type to shift an image such that objects move by (dx,dy) pixels. The shift is such that an object located at the coordinate (-dx, -dy) from the center of the image will be moved to the center of the image. So to move an object located 2 pixels right (dx) and 2 pixels up (dy) from the center to be at the center, use dx = -2, dy = -2.

Note that this does not treat the edges of the image, determined by the buffer width (lbuff) of the kernel and the size of shift. If you wish to treat the edges, you must pad the image by at least lbuff+abs(shift) pixels in each direction, and implement a strategy (zeros, mirror, wrap) prior to calling this function.

Template Parameters

arrOutT	is the Eigen-like array type of the output [will be resolved by compiler]
arrInT	is the Eigen-like array type of the input [will be resolved by compiler]
floatT1	is a floating point type [will be resolved by compiler]
floatT2	is a floating point type [will be resolved by compiler]
transformT	specifies the transformation to use [will be resolved by compiler]

Test:

Scenario: Verify direction and accuracy of various image shifts [test doc]

Parameters

[out]	transim	Will contain the shifted image. Will be allocated.
[in]	im	the image to be shifted.
[in]	dx	the amount to shift in the x direction
[in]	dy	the amount to shift in the y direction
[in]	trans	trans is the transformation to use

Definition at line 462 of file imageTransforms.hpp.

References mx::improc::imageShiftWP().

Referenced by mx::improc::ADIobservation< _realT, _derotFunctObj >::injectFake(), SCENARIO(), mx::AO::sim::turbLayer< _aoSystemT >::shift(), and mx::AO::sim::turbLayer< _aoSystemT >::shiftRandom().

imageShiftWP()

template<typename outputArrT , typename inputArrT >

void mx::improc::imageShiftWP	(	outputArrT &	out,
		inputArrT &	in,
		int	dx,
		int	dy,
		bool	wrap = true
	)

Shift an image by whole pixels with (optional) wrapping.

The output image can be smaller than the input image, in which case the wrapping (if enabled) still occurs for the input image, but only output images worth of pixels are actually shifted. This is useful, for instance, when propagating large turbulence phase screens where one only needs a small section at a time.

Template Parameters

outputArrT	is the eigen array type of the output [will be resolved by compiler]
inputArrT	is the eigen array type of the input [will be resolved by compiler]

Test:

Scenario: Verify direction and accuracy of various image shifts [test doc]

Parameters

[out]	out	contains the shifted image. Must be pre-allocated, but can be smaller than the in array.
[in]	in	the image to be shifted.
[in]	dx	the amount to shift in the x direction
[in]	dy	the amount to shift in the y direction
[in]	wrap	flag controlling whether or not to wrap around

Definition at line 294 of file imageTransforms.hpp.

Referenced by mx::improc::imageShift(), mx::improc::imageXCorrFFT< _realImageT >::refIm(), mx::AO::sim::turbLayer< _aoSystemT >::shift(), and mx::AO::sim::turbLayer< _aoSystemT >::shiftRandom().

imageShiftWPScale()

template<typename outputArrT , typename inputArrT , typename scaleArrT >

void mx::improc::imageShiftWPScale	(	outputArrT &	out,
		inputArrT &	in,
		scaleArrT &	scale,
		int	dx,
		int	dy
	)

Shift an image by whole pixels, wrapping around, with a scaling image applied to the shifted image.

The output image can be smaller than the input image, in which case the wrapping still occurs for the input image, but only output images worth of pixels are actually shifted. This is useful, for instance, when propagating large turbulence phase screens where one only needs a small section at a time.

The scaling is applied to the output image. The scale image must be the same size as the output image.

Template Parameters

outputArrT	is the eigen array type of the output [will be resolved by compiler]
inputArrT	is the eigen array type of the input [will be resolved by compiler]
scaleArrT	is the eigen array type of the scale image [will be resolved by compiler]

Parameters

[out]	out	contains the shifted image. Must be pre-allocated, but can be smaller than the in array.
[in]	in	the image to be shifted.
[in]	scale	image of scale values applied per-pixel to the output (shifted) image, same size as out
[in]	dx	the amount to shift in the x direction
[in]	dy	the amount to shift in the y direction

Definition at line 392 of file imageTransforms.hpp.