#include <WaveletTransform.hh>

Inheritance diagram for WaveletTransform:

Collaboration diagram for WaveletTransform:

Public Member Functions
	WaveletTransform (Image &img, FilterSet &fil)
virtual	~WaveletTransform (void)
virtual void	analysis (int steps)
virtual void	synthesis (int steps, int prevSteps=0)
void	expandImage (void)
void	restoreImage (void)
virtual void	where (area what, int subband, int &yoffs, int &xoffs, int &ysize, int &xsize) const
virtual void	where (int ypos, int xpos, int &subband, area &channel, int steps=-1) const
virtual void	mapPosition (int yOld, int xOld, int toSubband, area toChannel, int &yNew, int &xNew, int &sizeFactor, int steps=-1) const
virtual coeff	saverage (int subband, area channel) const
virtual coeff	aaverage (int subband, area channel) const
virtual coeff	sqvariance (int subband, area channel) const
virtual coeff	variance (int subband, area channel, bool abs=false) const
virtual coeff	sdeviation (int subband, area channel, bool abs=false) const
virtual coeff	ratio (int subband1, int subband2)
Image *	ll (int steps=-1)
Image *	hl (int steps=-1)
Image *	lh (int steps=-1)
Image *	hh (int steps=-1)
Image *	highMax (int steps=-1)
Image &	image (void) const
virtual Image *	subband (area what, int steps=-1)
virtual int	getSubband (int ypos, int xpos, int steps=-1)
int	getSubband (CoeffInformation &c, int steps=-1)
virtual area	getArea (int ypos, int xpos, int steps=-1)
area	getArea (CoeffInformation &c, int steps=-1)
virtual void	import (Image &img, area what, int steps=-1)
virtual void	fill (coeff value, area what, int steps=-1)
int	steps (void) const
Protected Member Functions
virtual void	doanalysis (int steps)=0
virtual void	dosynthesis (int steps, int prevSteps=0)=0
Protected Attributes
Wavelet *	m_wavelet
Image *	m_image
Image *	m_buffer
FilterSet *	m_filter
int	m_rows
int	m_cols
Private Member Functions
void	sanity (void)

Detailed Description

An abstract Wavelet Transform. The Transform transform is two-dimensional, it thus works on images.

Definition at line 26 of file WaveletTransform.hh.

Constructor & Destructor Documentation

WaveletTransform::WaveletTransform	(	Image &	img,
		FilterSet &	fil
	)

Constructor. Initializes references to Image and Wavelet.

Parameters:

img	the image
fil	the filter set for the wavelet

virtual WaveletTransform::~WaveletTransform ( void ) [virtual]

Destructor. Releases allocated objects.

Member Function Documentation

virtual coeff WaveletTransform::aaverage	(	int	subband,
		area	channel
	)		const `[virtual]`

Return the average color for a particular area in a subband inside the image. The coefficients' signedness will be discarded. This is a shortcut to the corresponding method in the Image class.

Parameters:

subband	the subband
channel	the area in that subband

Returns:: the average color

virtual void WaveletTransform::analysis ( int steps ) [virtual]

Perform a Wavelet transform on the image.

Exceptions:

invalid_argument image is not square

Parameters:

steps the number of transform steps

virtual void WaveletTransform::doanalysis ( int steps ) [protected, pure virtual]

Perform a Wavelet transform on the image.

Parameters:

steps the number of transform steps

Implemented in StandardTransform, and PyramidTransform.

virtual void WaveletTransform::dosynthesis	(	int	steps,
		int	prevSteps = `0`
	)		`[protected, pure virtual]`

Perform an inverse Wavelet transform on the image.

Parameters:

steps	the number of inverse transform steps
prevSteps	if greater than 0, the number of previous decomposition steps to assume. This is necessary to reconstruct images that are not square and have side lengths which are not powers of two.

Implemented in StandardTransform, and PyramidTransform.

void WaveletTransform::expandImage ( void )

Temporarily expand the image (square, side length a power of two) to make a transform possible.

virtual void WaveletTransform::fill	(	coeff	value,
		area	what,
		int	steps = `-1`
	)		`[virtual]`

Fill one of the subbands with one particular value.

Exceptions:

invalid_argument the imported image does not match the calculated subband size

Parameters:

value	the value to be inserted
what	the subband, out of LL, HL, LH, HH
steps	assume number of transform steps instead of those computed from the previous transforms

virtual area WaveletTransform::getArea	(	int	ypos,
		int	xpos,
		int	steps = `-1`
	)		`[virtual]`

Get the area of a given position in the image. This default method of calculation refers to the Pyramid transform and due to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

ypos	the position's row
xpos	the position's col
steps	the number of steps we assume the image to have been transformed

Returns:: the position's subband

area WaveletTransform::getArea	(	CoeffInformation &	c,
		int	steps = `-1`
	)		`[inline]`

Get the area of a given position in the image denoted by a {CoeffInformation} object. This default method of calculation refers to the Pyramid transform and due * to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

c	a reference the {CoeffInformation} object.
steps	the number of steps we assume the image to have been transformed

Returns:: the position's subband

Definition at line 276 of file WaveletTransform.hh.

References CoeffInformation::xpos(), and CoeffInformation::ypos().

virtual int WaveletTransform::getSubband	(	int	ypos,
		int	xpos,
		int	steps = `-1`
	)		`[virtual]`

Get the subband of a given position in the image. This default method of calculation refers to the Pyramid transform and due to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

ypos	the position's row
xpos	the position's col
steps	the number of steps we assume the image to have been transformed

Returns:: the position's subband

int WaveletTransform::getSubband	(	CoeffInformation &	c,
		int	steps = `-1`
	)		`[inline]`

Get the subband of a given position in the image denoted by a {CoeffInformation} object. This default method of calculation refers to the Pyramid transform and due * to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

c	a reference the {CoeffInformation} object.
steps	the number of steps we assume the image to have been transformed

Returns:: the position's subband

Definition at line 246 of file WaveletTransform.hh.

References CoeffInformation::xpos(), and CoeffInformation::ypos().

Image* WaveletTransform::hh ( int steps = -1 ) [inline]

Get a subpicture containing the HH subband.

Parameters:

steps assume number of transform steps instead of those computed from the previous transforms

Definition at line 199 of file WaveletTransform.hh.

References HH.

Image* WaveletTransform::highMax ( int steps = -1 )

Get an image consisting of the given subband's highpass components maxima.

Parameters:

steps assume number of transform steps instead of those computed from

Returns:: a new image object containing the maxima as coefficients

Image* WaveletTransform::hl ( int steps = -1 ) [inline]

Get a subpicture containing the HL subband.

Parameters:

steps assume number of transform steps instead of those computed from the previous transforms

Definition at line 189 of file WaveletTransform.hh.

References HL.

Image& WaveletTransform::image ( void ) const [inline]

Get a reference to the image

Returns:: the reference to the image

Definition at line 210 of file WaveletTransform.hh.

virtual void WaveletTransform::import	(	Image &	img,
		area	what,
		int	steps = `-1`
	)		`[virtual]`

Import a subpicture containing one of the subbands.

Exceptions:

invalid_argument the imported image does not match the calculated subband size

Parameters:

img	the image containing the subband
what	the subband, out of LL, HL, LH, HH
steps	assume number of transform steps instead of those computed from the previous transforms

Image* WaveletTransform::lh ( int steps = -1 ) [inline]

Get a subpicture containing the LH subband.

Parameters:

steps assume number of transform steps instead of those computed from the previous transforms

Definition at line 194 of file WaveletTransform.hh.

References LH.

Image* WaveletTransform::ll ( int steps = -1 ) [inline]

Get a subpicture containing the LL subband.

Parameters:

steps assume number of transform steps instead of those computed from the previous transforms

Definition at line 184 of file WaveletTransform.hh.

References LL.

virtual void WaveletTransform::mapPosition	(	int	yOld,
		int	xOld,
		int	toSubband,
		area	toChannel,
		int &	yNew,
		int &	xNew,
		int &	sizeFactor,
		int	steps = `-1`
	)		const `[virtual]`

Identify channel and subband of a given position in the transformed image and map it to another subband/channel. This default method of calculation refers to the Pyramid transform and due to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

yOld	the source location's row.
xOld	the source location's column.
toSubband	the target subband
toChannel	the target channel
yNew	where the target location's row gets written to.
xNew	where the target location's column gets written to.
sizeFactor	where the target sizeFactor gets written to.
steps	the current number of transform steps to be assumed (-1 for automatic determination)

virtual coeff WaveletTransform::ratio	(	int	subband1,
		int	subband2
	)		`[virtual]`

Return the ratio resulting from the zerotree relationship between two subbands (e.g. 1:4 from subband 1 to 2). This method depends on the geometry associated with the `Pyramid' decomposition. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

subband1	the first subband
subband2	the second subband

Returns:: the ratio, e.g. 4 for 1:4

void WaveletTransform::restoreImage ( void )

Restore the temporarily expanded image back to original size.

void WaveletTransform::sanity ( void ) [private]

Resize m_buffer for transforms if image dimensions have changed

virtual coeff WaveletTransform::saverage	(	int	subband,
		area	channel
	)		const `[virtual]`

Return the average color for a particular area in a subband inside the image. The coefficients' signedness will be considered. This is a shortcut to the corresponding method in the Image class.

Parameters:

subband	the subband
channel	the area in that subband

Returns:: the average color

virtual coeff WaveletTransform::sdeviation	(	int	subband,
		area	channel,
		bool	abs = `false`
	)		const `[virtual]`

Returns the standard deviation for a particular area in a subband inside the image. This is a shortcut to the corresponding method in the Image class.

Parameters:

subband	the subband
channel	the area in that subband
abs	true if signs are discarded

Returns:: the standard deviation.

virtual coeff WaveletTransform::sqvariance	(	int	subband,
		area	channel
	)		const `[virtual]`

Return the square variance for a particular area in a subband inside the image. This is a shortcut to the corresponding method in the Image class.

Parameters:

subband	the subband
channel	the area in that subband

Returns:: the square variance

int WaveletTransform::steps ( void ) const [inline]

Return the current number of decomposition steps

Returns:: the number of decomposition steps

Definition at line 306 of file WaveletTransform.hh.

virtual Image* WaveletTransform::subband	(	area	what,
		int	steps = `-1`
	)		`[virtual]`

Get a subpicture containing one of the subbands.

Parameters:

what	the subband, out of LL, HL, LH, HH
steps	assume number of transform steps instead of those computed from the previous transforms

virtual void WaveletTransform::synthesis	(	int	steps,
		int	prevSteps = `0`
	)		`[virtual]`

Perform an inverse Wavelet transform on the image.

Exceptions:

invalid_argument image is not square

Parameters:

steps	the number of inverse transform steps
prevSteps	if greater than 0, the number of previous decomposition steps to assume. This is necessary to reconstruct images that are not square and have side lengths which are not powers of two.

virtual coeff WaveletTransform::variance	(	int	subband,
		area	channel,
		bool	abs = `false`
	)		const `[virtual]`

Return the variance for a particular area in a subband inside the image. This is a shortcut to the corresponding method in the Image class.

Parameters:

subband	the subband
channel	the area in that subband
abs	true if signs are discarded

Returns:: the square variance

virtual void WaveletTransform::where	(	area	what,
		int	subband,
		int &	yoffs,
		int &	xoffs,
		int &	ysize,
		int &	xsize
	)		const `[virtual]`

Identify position and size of a given channel in a given subband. This default method of calculation refers to the Pyramid transform and due to the lack of alternatives also to the Standard transform. Decomposition schemes using a different geometries, like e.g. the Packet transform may have to implement their own version of that method.

Parameters:

what	the channel we want
subband	the subband we are looking at
yoffs	the returned row
xoffs	the returned col
ysize	the returned number of rows
xsize	the returned number of cols

virtual void WaveletTransform::where	(	int	ypos,
		int	xpos,
		int &	subband,
		area &	channel,
		int	steps = `-1`
	)		const `[virtual]`

Identify channel and subband of a given position in the transformed image. This method depends on the `where' method and thus uses the geometry associated with the `Pyramid' decomposition.

Parameters:

ypos	the location's row.
xpos	the location's column.
subband	where the calculated subband gets written to.
channel	where the calculated channel gets written to.
steps	the current number of transform steps to be assumed (-1 for automatic determination)