CudaSegment Class Reference

CUDA Segmentation class. More...

Public Member Functions
void	CopyProbsToDevice (Image< float > **probs_h)
	Copy probabilities to GPU.
	CudaSegment (int width, int height)
	Constructor.
void	Execute (FgBgSegment &segment, Image< uchar > &cimg, Image< float > &dimg, int numFigures, bool download=true)
	Execute Segmentation.
	~CudaSegment ()
Static Public Attributes
static const int	max_figures = 8
	Maximum number of foreground objects.
Private Member Functions
float	ComputeBeta (uchar *image_d)
	Compute the surface parameter beta.
void	CreateHistograms (FgBgSegment &segment)
	Create all histograms.
void	SetGradientCosts (uchar *image_d, float beta, float gamma)
	Set gradient costs.
Private Attributes
uchar *	allocated_d
	Memory allocation block pointer.
float *	beliefs_d [max_figures+2]
	Posterior label beliefs.
uint *	bins_d
	Colour histogram bin indices.
uchar *	cimd_d
	Colour image.
float *	costh_d
	Horizontal Gradient costs.
float *	costv_d
	Vertical Gradient costs.
float *	dimd_d
	Disparity map.
int	height
	Image height.
float *	priors_d [max_figures+2]
	Prior label probabilities.
float *	probs_d [max_figures+2]
	Posterior label probabilities.
int *	sums_d
	Partial gradient magnitude sums.
uchar *	vimd_d
	Gray-level image.
int	width
	Image width.

Detailed Description

CUDA Segmentation class.

Definition at line 43 of file cudasegment.h.

Constructor.

Segmentation

Parameters:

	width	Width of the Image
	height	Height of the Image

CudaSegment::~CudaSegment ( )

float CudaSegment::ComputeBeta ( uchar * image_d ) [private]

Compute the surface parameter beta.

Compute the surface parameter beta

Parameters:

image_d

Disparity image

void CudaSegment::CopyProbsToDevice ( Image< float > ** probs_h )

Copy probabilities to GPU.

void CudaSegment::CreateHistograms ( FgBgSegment & segment ) [private]

Create all histograms.

Create all histograms for foreground, background and flat surface hypotheses. Currently not used because works only for smal histograms.

Parameters:

segment

Segmentation object

Execute Segmentation.

Execute Segmentation on GPU

Parameters:

	segment	Segmentation Object
	cimg	RGB input Image
	dimg	Disparity Image
	numFigures	Number of foreground objects
	download	Whether or not to copy new image data onto GPU