asrl::GpuSurfOctave Class Reference

A class that reserves memory on the GPU for processing the interest point operator. More...

#include <GpuSurfOctave.hpp>

Public Member Functions
	GpuSurfOctave ()
	GpuSurfOctave (int img_width, int img_height, float l1, float l2, float l3, float l4, float edge_scale, float base_scale, int octave, int baseStep, int nIntervals)
void	init (int img_width, int img_height, float l1, float l2, float l3, float l4, float edge_scale, float base_scale, int octave, int baseStep, int nIntervals)
	operator SurfOctaveParameters ()
float	mask_width ()
float	mask_height ()
bool	valid ()
int	border ()
int	height ()
int	width ()
int	intervals ()
int	step ()
const float *	scales ()
int	stride ()
Private Attributes
float	m_mask_width
	Dxx/Dyy mask half-width.
float	m_mask_height
	Dxx/Dyy mask half-height.
float	m_l1
	Filter mask parameter l1.
float	m_l2
	Filter mask parameter l2.
float	m_l3
	Filter mask parameter l3.
float	m_l4
	Filter mask parameter l4.
float	m_edge_scale
	Relative filter response weighting used as: response = DxxDyy - edge_scaleDxy*Dxy.
bool	m_valid
	Flag for whether or not a valid Hessian buffer can be created.
int	m_border
	Size of the border to ensure filter lookups do not exceed the image boundaries.
int	m_width
	Hessian buffer width.
int	m_height
	Hessian buffer height.
int	m_intervals
	Number of intervals in the octave.
int	m_step
	Step size (in pixels).
float	m_scales [ASRL_SURF_MAX_INTERVALS]
	The Hessian buffers for both the GPU and CPU.

Detailed Description

A class that reserves memory on the GPU for processing the interest point operator.

This class sets up the memory and computes the parameters for each octave for the Fast Hessian computation. Custom box filter dimensions can be specified using the l1-l4 mask parameters, which are utilized as illustrated in the following diagram.

Dxx/Dyy and Dxy filter mask parameters

Definition at line 59 of file GpuSurfOctave.hpp.

Constructor & Destructor Documentation

asrl::GpuSurfOctave::GpuSurfOctave ( )

The default constructor. This is included so the octave can be stored in a standard library container.

Definition at line 35 of file GpuSurfOctave.cpp.

asrl::GpuSurfOctave::GpuSurfOctave	(	int	img_width,
		int	img_height,
		float	l1,
		float	l2,
		float	l3,
		float	l4,
		float	edge_scale,
		float	base_scale,
		int	octave,
		int	baseStep,
		int	nIntervals
	)

The constructor with input parameters.

Parameters:

	img_width	The image width in pixels
	img_height	The image height in pixels
	l1	Filter mask parameter l1
	l2	Filter mask parameter l2
	l3	Filter mask parameter l3
	l4	Filter mask parameter l4
	edge_scale	Relative filter response weighting used as: response = DxxDyy - edge_scaleDxy*Dxy
	base_scale	Base scale for the first octave
	octave	Octave number (zero-indexed)
	baseStep	Initial step size (for the first octave) in pixels
	nIntervals	Number of intervals in the octave

Definition at line 39 of file GpuSurfOctave.cpp.

Member Function Documentation

int asrl::GpuSurfOctave::border ( ) [inline]

Returns:: The size of the border (in pixels) such that no evaluated filter exceeds the bounds of the image.

Definition at line 111 of file GpuSurfOctave.hpp.

int asrl::GpuSurfOctave::height ( ) [inline]

Returns:: The height of the Hessian buffer.

Definition at line 116 of file GpuSurfOctave.hpp.

void asrl::GpuSurfOctave::init	(	int	img_width,
		int	img_height,
		float	l1,
		float	l2,
		float	l3,
		float	l4,
		float	edge_scale,
		float	base_scale,
		int	octave,
		int	baseStep,
		int	nIntervals
	)