lti::gaborKernelSep< T > Class Template Reference

two-dimensional separable gabor filter kernel More...

#include <ltiGaborKernels.h>

Inheritance diagram for lti::gaborKernelSep< T >:

[legend]

Collaboration diagram for lti::gaborKernelSep< T >:

[legend]

List of all members.

Public Member Functions
	gaborKernelSep (const int &size, const double &theVariance, const double &frequency, const double &angle)
void	generate (const int &size, const double &theVariance, const double &frequency, const double &angle)

---

Detailed Description

template<class T>
class lti::gaborKernelSep< T >

two-dimensional separable gabor filter kernel

The Gabor kernels are defined by the following equation:

The index m denotes the "band" of the filter (which variance it uses) and the index n the direction of the filter.

This kernel can be expressed also as:

This two-filter-pair separation property is used in this implementation.

Due to the nature of this filter, avoid its use with unsigned byte channels! (no lti::channel8, only lti::channel).

Example:

      // the channel to be filtered:
      lti::channel data;

      // ... initialize channel here ...

      // gabor filter kernel with 7 elements, a variance of 1.0, frequency
      // of 0.5 and 45°
      lti::gaborKernelSep<channel::value_type> kernel(3,1.0,0.5,45*Pi/180.0);

      lti::convolution filter;                        // convolution operator
      lti::convolution::parameters param;             // parameters
      param.setKernel(kernel);                        // use the gauss kernel
      filter.setParameters(param);                    // set parameters

      // filter the vector and leave the result there too

      filter.apply(data);

Generally, it is faster to use the separable kernel for kernel dimension bigger or equal han 5x5.

---

Constructor & Destructor Documentation

template<class T>

lti::gaborKernelSep< T >::gaborKernelSep	(	const int &	size,
		const double &	theVariance,
		const double &	frequency,
		const double &	angle
	)

constructor

Parameters:

	size	is the number of elements pro dimension (i.e. the filter kern is a 'size x size' kernel!)
	theVariance	variance of the kernel
	frequency	the angular frequency to be modulated with (in radians per line)
	angle	the angle to modulate the gaussian function (in radians)

---

Member Function Documentation

template<class T>

void lti::gaborKernelSep< T >::generate	(	const int &	size,
		const double &	theVariance,
		const double &	frequency,
		const double &	angle
	)

initialize this kernel with the specified values

Parameters:

	size	size of the kernel in one dimension
	theVariance	variance of the kernel
	frequency	the angular frequency to be modulated with (in radians per line)
	angle	the angle to modulate the gaussian function (in radians)

---

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

• ltiGaborKernels.h