Class for smoothing the images before running BP. Smoothing image always uses float data type. More...

#include <SmoothImage.h>

Inheritance diagram for SmoothImage:

Collaboration diagram for SmoothImage:

Public Member Functions
	SmoothImage (const ParallelParams &parallel_params)

virtual void	operator() (const BpImage< unsigned int > &in_image, float sigma, float *smoothed_image) const =0
	Function to use the image filter to apply a guassian filter to the a single image. Input images have each pixel stored as an unsigned int (value between 0 and 255 assuming 8-bit grayscale image used). Output filtered images have each pixel stored as a float after the image has been smoothed with a Gaussian filter of sigma. Mask is normalized so it integrates to one. More...

Protected Member Functions
std::vector< float >	MakeFilter (float sigma) const
	Create a Gaussian filter from a sigma value. More...

Protected Attributes
const ParallelParams &	parallel_params_
	Parallel parameters to use parallel operations (number of threads on CPU / thread block config in CUDA) More...

Detailed Description

Class for smoothing the images before running BP. Smoothing image always uses float data type.

Definition at line 49 of file SmoothImage.h.

Constructor & Destructor Documentation

◆ SmoothImage()

SmoothImage::SmoothImage ( const ParallelParams & parallel_params )

inlineexplicit

Definition at line 52 of file SmoothImage.h.

Member Function Documentation

◆ MakeFilter()

std::vector< float > SmoothImage::MakeFilter ( float sigma ) const

protected

Create a Gaussian filter from a sigma value.

Parameters

sigma

Returns: Generated Gaussian filter

Definition at line 30 of file SmoothImage.cpp.

Here is the caller graph for this function:

◆ operator()()

virtual void SmoothImage::operator()	(	const BpImage< unsigned int > &	in_image,
		float	sigma,
		float *	smoothed_image
	)		const

pure virtual

Function to use the image filter to apply a guassian filter to the a single image.
Input images have each pixel stored as an unsigned int (value between 0 and 255 assuming 8-bit grayscale image used).
Output filtered images have each pixel stored as a float after the image has been smoothed with a Gaussian filter of sigma.
Mask is normalized so it integrates to one.

Parameters

in_image
sigma
smoothed_image

Implemented in SmoothImageCUDA, and SmoothImageCPU.

Member Data Documentation

◆ parallel_params_

const ParallelParams& SmoothImage::parallel_params_

protected

Parallel parameters to use parallel operations (number of threads on CPU / thread block config in CUDA)

Definition at line 86 of file SmoothImage.h.

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

src/beliefprop/BpImageProcessing/SmoothImage.h
src/beliefprop/BpImageProcessing/SmoothImage.cpp

Public Member Functions

Protected Member Functions

Protected Attributes