Convolution is a mathematical operation on two functions f and g, producing a third function that is typically viewed as a modified version of one of the original functions. More...

#include <pcl/filters/convolution.h>

Public Types
enum	BORDERS_POLICY { BORDERS_POLICY_IGNORE = -1 , BORDERS_POLICY_MIRROR = 0 , BORDERS_POLICY_DUPLICATE = 1 }
	The borders policy available. More...
using	PointCloudIn = pcl::PointCloud<PointIn>
using	PointCloudInPtr = typename PointCloudIn::Ptr
using	PointCloudInConstPtr = typename PointCloudIn::ConstPtr
using	PointCloudOut = pcl::PointCloud<PointOut>
using	Ptr = shared_ptr< Convolution<PointIn, PointOut> >
using	ConstPtr = shared_ptr< const Convolution<PointIn, PointOut> >

Public Member Functions
	Convolution ()
	Constructor.
	~Convolution ()
	Empty destructor.
void	setInputCloud (const PointCloudInConstPtr &cloud)
	Provide a pointer to the input dataset.
void	setKernel (const Eigen::ArrayXf &kernel)
	Set convolving kernel.
void	setBordersPolicy (int policy)
	Set the borders policy.
int	getBordersPolicy ()
	Get the borders policy.
void	setDistanceThreshold (const float &threshold)
const float &	getDistanceThreshold () const
void	setNumberOfThreads (unsigned int nr_threads=0)
	Initialize the scheduler and set the number of threads to use.
void	convolveRows (PointCloudOut &output)
	Convolve a float image rows by a given kernel.
void	convolveCols (PointCloudOut &output)
	Convolve a float image columns by a given kernel.
void	convolve (const Eigen::ArrayXf &h_kernel, const Eigen::ArrayXf &v_kernel, PointCloudOut &output)
	Convolve point cloud with an horizontal kernel along rows then vertical kernel along columns : convolve separately.
void	convolve (PointCloudOut &output)
	Convolve point cloud with same kernel along rows and columns separately.

Protected Member Functions
void	convolve_rows (PointCloudOut &output)
	convolve rows and ignore borders
void	convolve_cols (PointCloudOut &output)
	convolve cols and ignore borders
void	convolve_rows_mirror (PointCloudOut &output)
	convolve rows and mirror borders
void	convolve_cols_mirror (PointCloudOut &output)
	convolve cols and mirror borders
void	convolve_rows_duplicate (PointCloudOut &output)
	convolve rows and duplicate borders
void	convolve_cols_duplicate (PointCloudOut &output)
	convolve cols and duplicate borders
void	initCompute (PointCloudOut &output)
	init compute is an internal method called before computation
void	makeInfinite (PointOut &p)
void	makeInfinite (pcl::RGB &p)

Protected Attributes
unsigned int	threads_
	The number of threads the scheduler should use.

Detailed Description

template<typename PointIn, typename PointOut>
class pcl::filters::Convolution< PointIn, PointOut >

Convolution is a mathematical operation on two functions f and g, producing a third function that is typically viewed as a modified version of one of the original functions.

see http://en.wikipedia.org/wiki/Convolution.

The class provides rows, column and separate convolving operations of a point cloud. Columns and separate convolution is only allowed on organised point clouds.

When convolving, computing the rows and cols elements at 1/2 kernel width distance from the borders is not defined. We allow for 3 policies:

Ignoring: elements at special locations are filled with zero (default behaviour)
Mirroring: the missing rows or columns are obtained through mirroring
Duplicating: the missing rows or columns are obtained through duplicating

Author: Nizar Sallem

Definition at line 72 of file convolution.h.

Member Typedef Documentation

◆ ConstPtr

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::ConstPtr = shared_ptr< const Convolution<PointIn, PointOut> >

Definition at line 80 of file convolution.h.

◆ PointCloudIn

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::PointCloudIn = pcl::PointCloud<PointIn>

Definition at line 75 of file convolution.h.

◆ PointCloudInConstPtr

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::PointCloudInConstPtr = typename PointCloudIn::ConstPtr

Definition at line 77 of file convolution.h.

◆ PointCloudInPtr

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::PointCloudInPtr = typename PointCloudIn::Ptr

Definition at line 76 of file convolution.h.

◆ PointCloudOut

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::PointCloudOut = pcl::PointCloud<PointOut>

Definition at line 78 of file convolution.h.

◆ Ptr

template<typename PointIn, typename PointOut>

using pcl::filters::Convolution< PointIn, PointOut >::Ptr = shared_ptr< Convolution<PointIn, PointOut> >

Definition at line 79 of file convolution.h.

Member Enumeration Documentation

◆ BORDERS_POLICY

template<typename PointIn, typename PointOut>

enum pcl::filters::Convolution::BORDERS_POLICY

The borders policy available.

Enumerator
BORDERS_POLICY_IGNORE
BORDERS_POLICY_MIRROR
BORDERS_POLICY_DUPLICATE

Definition at line 84 of file convolution.h.

Constructor & Destructor Documentation

◆ Convolution()

template<typename PointIn, typename PointOut>

pcl::filters::Convolution< PointIn, PointOut >::Convolution ( )

Constructor.

Definition at line 54 of file convolution.hpp.

References BORDERS_POLICY_IGNORE, and threads_.

◆ ~Convolution()

template<typename PointIn, typename PointOut>

pcl::filters::Convolution< PointIn, PointOut >::~Convolution ( )

inline

Empty destructor.

Definition at line 93 of file convolution.h.

Member Function Documentation

◆ convolve() [1/2]

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve	(	const Eigen::ArrayXf &	h_kernel,
		const Eigen::ArrayXf &	v_kernel,
		PointCloudOut &	output )

inline

Convolve point cloud with an horizontal kernel along rows then vertical kernel along columns : convolve separately.

Parameters

[in]	h_kernel	kernel for convolving rows
[in]	v_kernel	kernel for convolving columns
[out]	output	the convolved cloud

Note: if output doesn't fit in input i.e. output.rows () < input.rows () or output.cols () < input.cols () then output is resized to input sizes.

Definition at line 135 of file convolution.hpp.

References convolveCols(), convolveRows(), setInputCloud(), and setKernel().

Referenced by pcl::ColorGradientModality< PointInT >::processInputData().

◆ convolve() [2/2]

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve ( PointCloudOut & output )

inline

Convolve point cloud with same kernel along rows and columns separately.

Parameters

[out] output the convolved cloud

Note: if output doesn't fit in input i.e. output.rows () < input.rows () or output.cols () < input.cols () then output is resized to input sizes.

Definition at line 156 of file convolution.hpp.

References convolveCols(), convolveRows(), and setInputCloud().

◆ convolve_cols()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_cols ( PointCloudOut & output )

protected

convolve cols and ignore borders

Definition at line 421 of file convolution.hpp.

References makeInfinite().

Referenced by convolveCols().

◆ convolve_cols_duplicate()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_cols_duplicate ( PointCloudOut & output )

protected

convolve cols and duplicate borders

Definition at line 467 of file convolution.hpp.

Referenced by convolveCols().

◆ convolve_cols_mirror()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_cols_mirror ( PointCloudOut & output )

protected

convolve cols and mirror borders

Definition at line 514 of file convolution.hpp.

Referenced by convolveCols().

◆ convolve_rows()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_rows ( PointCloudOut & output )

protected

convolve rows and ignore borders

Definition at line 281 of file convolution.hpp.

References makeInfinite().

Referenced by convolveRows().

◆ convolve_rows_duplicate()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_rows_duplicate ( PointCloudOut & output )

protected

convolve rows and duplicate borders

Definition at line 327 of file convolution.hpp.

Referenced by convolveRows().

◆ convolve_rows_mirror()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolve_rows_mirror ( PointCloudOut & output )

protected

convolve rows and mirror borders

Definition at line 374 of file convolution.hpp.

Referenced by convolveRows().

◆ convolveCols()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolveCols ( PointCloudOut & output )

inline

Convolve a float image columns by a given kernel.

Parameters

[out] output the convolved image

Note: if output doesn't fit in input i.e. output.rows () < input.rows () or output.cols () < input.cols () then output is resized to input sizes.

Definition at line 115 of file convolution.hpp.

References BORDERS_POLICY_DUPLICATE, BORDERS_POLICY_IGNORE, BORDERS_POLICY_MIRROR, convolve_cols(), convolve_cols_duplicate(), convolve_cols_mirror(), and initCompute().

Referenced by convolve(), and convolve().

◆ convolveRows()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::convolveRows ( PointCloudOut & output )

inline

Convolve a float image rows by a given kernel.

Parameters

[out] output the convolved cloud

Note: if output doesn't fit in input i.e. output.rows () < input.rows () or output.cols () < input.cols () then output is resized to input sizes.

Definition at line 95 of file convolution.hpp.

References BORDERS_POLICY_DUPLICATE, BORDERS_POLICY_IGNORE, BORDERS_POLICY_MIRROR, convolve_rows(), convolve_rows_duplicate(), convolve_rows_mirror(), and initCompute().

Referenced by convolve(), and convolve().

◆ getBordersPolicy()

template<typename PointIn, typename PointOut>

int pcl::filters::Convolution< PointIn, PointOut >::getBordersPolicy ( )

inline

Get the borders policy.

Definition at line 110 of file convolution.h.

◆ getDistanceThreshold()

template<typename PointIn, typename PointOut>

const float & pcl::filters::Convolution< PointIn, PointOut >::getDistanceThreshold ( ) const

inline

Returns: the distance threshold

Definition at line 122 of file convolution.h.

◆ initCompute()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::initCompute ( PointCloudOut & output )

protected

init compute is an internal method called before computation

Parameters

[in] output

Exceptions

pcl::InitFailedException

Definition at line 64 of file convolution.hpp.

References BORDERS_POLICY_DUPLICATE, BORDERS_POLICY_IGNORE, BORDERS_POLICY_MIRROR, pcl::PointCloud< PointT >::height, pcl::PointCloud< PointT >::is_dense, pcl::PointCloud< PointT >::resize(), and pcl::PointCloud< PointT >::width.

Referenced by convolveCols(), and convolveRows().

◆ makeInfinite() [1/2]

void pcl::filters::Convolution< pcl::RGB, pcl::RGB >::makeInfinite ( pcl::RGB & p )

inlineprotected

Definition at line 275 of file convolution.hpp.

◆ makeInfinite() [2/2]

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::makeInfinite ( PointOut & p )

inlineprotected

Definition at line 224 of file convolution.h.

Referenced by convolve_cols(), and convolve_rows().

◆ setBordersPolicy()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::setBordersPolicy ( int policy )

inline

Set the borders policy.

Definition at line 107 of file convolution.h.

◆ setDistanceThreshold()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::setDistanceThreshold ( const float & threshold )

inline

Remarks: this is critical so please read it carefully. In 3D the next point in (u,v) coordinate can be really far so a distance threshold is used to keep us from ghost points. The value you set here is strongly related to the sensor. A good value for kinect data is 0.001. Default is std::numeric<float>::infinity ()

Parameters

[in] threshold maximum allowed distance between 2 juxtaposed points

Definition at line 119 of file convolution.h.

◆ setInputCloud()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::setInputCloud ( const PointCloudInConstPtr & cloud )

inline

Provide a pointer to the input dataset.

Parameters

cloud the const boost shared pointer to a PointCloud message

Remarks: Will perform a deep copy

Definition at line 99 of file convolution.h.

Referenced by convolve(), convolve(), and pcl::ColorGradientModality< PointInT >::processInputData().

◆ setKernel()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::setKernel ( const Eigen::ArrayXf & kernel )

inline

Set convolving kernel.

Parameters

[in] kernel convolving element

Definition at line 104 of file convolution.h.

Referenced by convolve(), and pcl::ColorGradientModality< PointInT >::processInputData().

◆ setNumberOfThreads()

template<typename PointIn, typename PointOut>

void pcl::filters::Convolution< PointIn, PointOut >::setNumberOfThreads ( unsigned int nr_threads = 0 )

inline

Initialize the scheduler and set the number of threads to use.

Parameters

nr_threads the number of hardware threads to use (0 sets the value back to automatic)

Definition at line 127 of file convolution.h.

References threads_.

Member Data Documentation

◆ threads_

template<typename PointIn, typename PointOut>

unsigned int pcl::filters::Convolution< PointIn, PointOut >::threads_

protected

The number of threads the scheduler should use.

Definition at line 221 of file convolution.h.

Referenced by Convolution(), and setNumberOfThreads().

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

pcl/filters/convolution.h
pcl/filters/impl/convolution.hpp

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Typedef Documentation

◆ ConstPtr

◆ PointCloudIn

◆ PointCloudInConstPtr

◆ PointCloudInPtr

◆ PointCloudOut

◆ Ptr

Member Enumeration Documentation

◆ BORDERS_POLICY

Constructor & Destructor Documentation

◆ Convolution()

◆ ~Convolution()

Member Function Documentation

◆ convolve() [1/2]

◆ convolve() [2/2]

◆ convolve_cols()

◆ convolve_cols_duplicate()

◆ convolve_cols_mirror()

◆ convolve_rows()

◆ convolve_rows_duplicate()

◆ convolve_rows_mirror()

◆ convolveCols()

◆ convolveRows()

◆ getBordersPolicy()

◆ getDistanceThreshold()

◆ initCompute()

◆ makeInfinite() [1/2]

◆ makeInfinite() [2/2]

◆ setBordersPolicy()

◆ setDistanceThreshold()

◆ setInputCloud()

◆ setKernel()

◆ setNumberOfThreads()

Member Data Documentation

◆ threads_