Utility struct for managing convolution kernel data. More...

#include <ConvolutionKernel.h>

Public Types
enum	fixedType { gauss3x3 = 0, gauss5x5 = 10, sobelX3x3 = 1, sobelX5x5 = 11, sobelY3x3 = 2, sobelY5x5 = 12, laplace3x3 = 3, laplace5x5 = 13, custom }
	this enum contains several predefined convolution kernel types More...
Public Member Functions
	ConvolutionKernel ()
	creates an empty null kernel
	ConvolutionKernel (const ConvolutionKernel &other)
	crates a copied kernel
	ConvolutionKernel (int *data, const utils::Size &size, int factor=1, bool deepCopy=true) throw (utils::InvalidSizeException)
	create an integer based kernel
	ConvolutionKernel (float *data, const utils::Size &size, bool deepCopy=true) throw (utils::InvalidSizeException)
	create a float valued kernel
	ConvolutionKernel (fixedType t, bool useFloats=false)
	create a fixed kernel (optionally as float kernel)
	~ConvolutionKernel ()
	Destructor.
ConvolutionKernel &	operator= (const ConvolutionKernel &other)
	Assignment operator.
void	toFloat ()
	converts internal idata to float data regarding 'factor'
void	toInt (bool force=false)
	converts fdata to idata if possible and/or if force flag is true
int	getDim () const
	returns kernels width * height
const utils::Size &	getSize () const
	returns the kernels size
int	getWidth () const
	returns the kernels width
int	getHeight () const
	returns the kernels height
bool	isFloat () const
	returns whether this kernel has float data
const float *	getFloatData () const
	returns the kernels float data pointer (const) (which may be 0)
float *	getFloatData ()
	returns the kernels float data pointer (which may be 0)
const int *	getIntData () const
	returns the kernels int data pointer (const) (which may be 0)
int *	getIntData ()
	returns the kernels int data pointer (which may be 0)
int	getFactor () const
	returns the kernels scaling factor (only used if its an integer kernel)
bool	isNull () const
	return whether idata or fdata is not null
fixedType	getFixedType () const
	returns the kernels fixed type (necessary for IPP-optimization)
void	detach ()
	ensures shallow copied data is copied deeply
Private Attributes
utils::Size	size
	associated size
float *	fdata
	float data pointer
int *	idata
	int data pointer
int	factor
	scaling factor for integer kernels
bool	isnull
	is already initialized
bool	owned
	is data owned
fixedType	ft
	fixed type set

Detailed Description

Utility struct for managing convolution kernel data.

This structure provides a shallow as well as deep wrapper for int- and float-pointers into ConvolutionKernel-instances. Convolution Kernels can be float- or integer valued.

copies of shallow copied instances are shallow!

Member Enumeration Documentation

enum icl::filter::ConvolutionKernel::fixedType

this enum contains several predefined convolution kernel types

kernelSobleX

The sobel x filter is a combined filter. It performs a symmetrical border filter operation in x-direction, followed by a smoothing operation in y-direction:

                          ---------     ---        
                         | 1  0 -1 |   | 1 |    ---------
          kernelSobelX = | 2  0 -2 | = | 2 | * | 1  0 -1 |
                         | 1  0 -1 |   | 1 |    ---------
                          ---------     ---

kernelSobelY

The sobel y filter is essentially equal to the sobel y filter. The border detection will run in y-direction, and the smoothing x-direction.

                          ---------     ---        
                         | 1  2  1 |   | 1 |    ---------
          kernelSobelY = | 0  0  0 | = | 0 | * | 1  2  1 |
                         |-1 -2 -1 |   |-1 |    ---------
                          ---------     ---

kernelGauss3x3

This is a 3x3-Pixel approximation of a 2D Gaussian. It is separable into smoothing filters in x- and y-direction.

                                   ---------     ---        
                                  | 1  2  1 |   | 1 |    ---------
          kernelGauss3x3 = 1/16 * | 2  4  2 | = | 2 | * | 1  2  1 | * (1/4) *(1/4) 
                                  | 1  2  1 |   | 1 |    ---------
                                   ---------     ---

kernelGauss5x5

This is a 5x5-Pixel approximation of a 2D Gaussian. It is separable into smoothing filters in x- and y-direction.

                                    ----------------- 
                                   | 2   7  12  7  2 |
                                   | 7  31  52 31  7 |
          kernelGauss5x5 = 1/571 * |12  52 127 52 12 |
                                   | 7  31  52 31  7 |
                                   | 2   7  12  7  2 |
                                    -----------------

kernelLaplace

The Laplacian kernel is a discrete approximation of the 2nd derivation of a 2D function.

                           --------- 
                          | 1  1  1 |
          kernelLaplace = | 1 -8  1 |
                          | 1  1  1 |
                           ---------

Enumerator:

gauss3x3	3x3 approximation of a Gaussian
gauss5x5	5x5 approximation of a Gaussian
sobelX3x3	3x3 sobel x filter
sobelX5x5	5x5 sobel x filter
sobelY3x3	3x3 sobel y filter
sobelY5x5	5x5 sobel y filter
laplace3x3	3x3 approximation of the 2nd derivation
laplace5x5	5x5 approximation of the 2nd derivation
custom

Constructor & Destructor Documentation

icl::filter::ConvolutionKernel::ConvolutionKernel ( )

creates an empty null kernel

icl::filter::ConvolutionKernel::ConvolutionKernel ( const ConvolutionKernel & other )

crates a copied kernel

shallow wrapped kernels are copied shallowly !

icl::filter::ConvolutionKernel::ConvolutionKernel	(	int *	data,
		const utils::Size &	size,
		int	factor = `1`,
		bool	deepCopy = `true`
	)		throw (utils::InvalidSizeException)

create an integer based kernel

factor can be used to rescale convolution results by given factor: Result(x,y) = ( src *(x,y)* kernel ) / factor where *(x,y)* is a convolution at location (x,y)

icl::filter::ConvolutionKernel::ConvolutionKernel	(	float *	data,
		const utils::Size &	size,
		bool	deepCopy = `true`
	)		throw (utils::InvalidSizeException)

create a float valued kernel

icl::filter::ConvolutionKernel::ConvolutionKernel	(	fixedType	t,
		bool	useFloats = `false`
	)

create a fixed kernel (optionally as float kernel)

icl::filter::ConvolutionKernel::~ConvolutionKernel ( )

Destructor.

Member Function Documentation

void icl::filter::ConvolutionKernel::detach ( )

ensures shallow copied data is copied deeply

int icl::filter::ConvolutionKernel::getDim ( ) const [inline]

returns kernels width * height

int icl::filter::ConvolutionKernel::getFactor ( ) const [inline]

returns the kernels scaling factor (only used if its an integer kernel)

fixedType icl::filter::ConvolutionKernel::getFixedType ( ) const [inline]

returns the kernels fixed type (necessary for IPP-optimization)

const float* icl::filter::ConvolutionKernel::getFloatData ( ) const [inline]

returns the kernels float data pointer (const) (which may be 0)

float* icl::filter::ConvolutionKernel::getFloatData ( ) [inline]

returns the kernels float data pointer (which may be 0)

int icl::filter::ConvolutionKernel::getHeight ( ) const [inline]

returns the kernels height

const int* icl::filter::ConvolutionKernel::getIntData ( ) const [inline]

returns the kernels int data pointer (const) (which may be 0)

int* icl::filter::ConvolutionKernel::getIntData ( ) [inline]

returns the kernels int data pointer (which may be 0)

const utils::Size& icl::filter::ConvolutionKernel::getSize ( ) const [inline]

returns the kernels size

int icl::filter::ConvolutionKernel::getWidth ( ) const [inline]

returns the kernels width

bool icl::filter::ConvolutionKernel::isFloat ( ) const [inline]

returns whether this kernel has float data

bool icl::filter::ConvolutionKernel::isNull ( ) const [inline]

return whether idata or fdata is not null

ConvolutionKernel& icl::filter::ConvolutionKernel::operator= ( const ConvolutionKernel & other )

Assignment operator.

Shallow-wrapped kernels are copied shallowly

void icl::filter::ConvolutionKernel::toFloat ( )

converts internal idata to float data regarding 'factor'

void icl::filter::ConvolutionKernel::toInt ( bool force = false )

converts fdata to idata if possible and/or if force flag is true

'possible' means that all float values do not show decimals

Member Data Documentation

int icl::filter::ConvolutionKernel::factor [private]

scaling factor for integer kernels

float* icl::filter::ConvolutionKernel::fdata [private]

float data pointer

fixedType icl::filter::ConvolutionKernel::ft [private]

fixed type set

int* icl::filter::ConvolutionKernel::idata [private]

int data pointer

bool icl::filter::ConvolutionKernel::isnull [private]

is already initialized

bool icl::filter::ConvolutionKernel::owned [private]

is data owned

utils::Size icl::filter::ConvolutionKernel::size [private]

associated size

The documentation for this struct was generated from the following file:

/home/jenkins/workspace/icl-manual-trunk/icl-manual/ICLFilter/src/ICLFilter/ConvolutionKernel.h

Public Types

Public Member Functions

Private Attributes