Functions
template<template< class D > class C>
float	ClusterMatchingError (const Vector< C< int > > &sim_label)
	Computes and returns the minimum error of corresponding clusters between pairs of containers. e.g. Given a ground truth labeled image and a test labeled image, this function tries to maximize the matching of cluster labels in order to achieve the minimum error. If both images have the same clusters, but with different labels assigned to them, the error will be zero. The error is the frequency of incorrect labeled pixels. More...

template<class D >
size_t	HierarchicalClustering (Feature< D > &feature, float scale_min, float scale_max)
	Computes OPF clustering based only on the feature space, using sampling and a graph. It tries to find the best scale for each cluster. More...

template<class D , class GRAPH_ADJACENCY >
size_t	HierarchicalClustering (Feature< D > &feature, float scale_min, float scale_max, HierarchicalGraph< GRAPH_ADJACENCY > &graph)

template<class D >
size_t	SpacialClustering (const Feature< D > &feature, Image< int > &label, const Adjacency &adjacency, float intensity_fraction)
	Computes OPF clustering based on the feature and image space, using complete graph. More...

template<class D >
float	MaxWeight (const Feature< D > &feature, Image< int > &label, const Adjacency &adjacency, float intensity_fraction)
	Computes PDF for 'neighbors' neighbours. More...

template<class D >
void	MaxWeightThread (const Feature< D > &feature, Image< int > &label, const Adjacency &adjacency, float &max_distance, size_t thread, size_t total_threads)
	Computes PDF for 'neighbors' neighbours. More...

template<class D >
float	PDF (const Feature< D > &feature, const Adjacency &adjacency, Image< float > &density, float max_weight)
	Computes PDF in spatial and spectral domains. More...

template<class D >
float	PDFThread (const Feature< D > &feature, const Adjacency &adjacency, Image< float > &density, float sigma, float &max_dens_diff, size_t thread, size_t total_threads)
	Computes PDF in spatial and spectral domains. More...

template<class D >
size_t	SpectralClustering (Feature< D > &feature, float scale_min, float scale_max)
	Computes OPF clustering based only on the feature space, using sampling and a graph. More...

template<class D , class GRAPH_ADJACENCY >
size_t	SpectralClustering (Feature< D > &feature, float scale_min, float scale_max, Graph< GRAPH_ADJACENCY > &graph)

template<class D >
size_t	SpectralClustering (Feature< D > &feature, size_t clusters)
	Computes OPF clustering based only on the feature space, using sampling and k-nn arcs. More...

template<class D , class GRAPH_ADJACENCY >
size_t	SpectralClustering (Feature< D > &feature, size_t clusters, Graph< GRAPH_ADJACENCY > &graph)

Function Documentation

template<template< class D > class C>

float Bial::OPF::ClusterMatchingError ( const Vector< C< int > > & sim_label )

Computes and returns the minimum error of corresponding clusters between pairs of containers. e.g. Given a ground truth labeled image and a test labeled image, this function tries to maximize the matching of cluster labels in order to achieve the minimum error. If both images have the same clusters, but with different labels assigned to them, the error will be zero. The error is the frequency of incorrect labeled pixels.

Date: 2014/Apr/30

Parameters

sim_label Vector of labels.

Returns: The minimum error of corresponding clusters between pairs of containers (i.e. images, vectors).

Warning: none.

template<class D >

size_t Bial::OPF::HierarchicalClustering	(	Feature< D > &	feature,
		float	scale_min,
		float	scale_max
	)

Computes OPF clustering based only on the feature space, using sampling and a graph. It tries to find the best scale for each cluster.

Date: 2014/Nov/04

Parameters

feature	Feature vector.
scale_min,scale_max	minimum and maximum scale. From 0.0 to 1.0.
graph	the graph used to compute OPF. Default: KnnGraph for function with no parameter.

Returns: Number of clusters.

Warning: scale_min must be smaller than of equal to scale_min.

template<class D , class GRAPH_ADJACENCY >

size_t Bial::OPF::HierarchicalClustering	(	Feature< D > &	feature,
		float	scale_min,
		float	scale_max,
		HierarchicalGraph< GRAPH_ADJACENCY > &	graph
	)

template<class D >

float Bial::OPF::MaxWeight	(	const Feature< D > &	feature,
		Image< int > &	label,
		const Adjacency &	adjacency,
		float	intensity_fraction
	)

Computes PDF for 'neighbors' neighbours.

Date: 2012/Nov/26

Parameters

feature	Feature vector.
label	label map. May be a std random access container, Bial::Image, or Bial::Matrix class.
adjacency	adjacency relation for spacial constraints.
intensity_fraction	fraction from 0.0 to 1.0 of the maximum intensity set for adjacent pixels.

Returns: The bucket size for the IFT queue and PathFunction.

Warning: none.

template<class D >

void Bial::OPF::MaxWeightThread	(	const Feature< D > &	feature,
		Image< int > &	label,
		const Adjacency &	adjacency,
		float &	max_distance,
		size_t	thread,
		size_t	total_threads
	)

Computes PDF for 'neighbors' neighbours.

Date: 2013/Dec/10

Parameters

feature	Feature vector.
label	label map. May be a std random access container, Bial::Image, or Bial::Matrix class.
adjacency	adjacency relation for spacial constraints.
max_distance	Maximum distance among pixels.
thread	Thread number.
total_threads	Number of threads.

Returns: The bucket size for the IFT queue and PathFunction.

Warning: none.

template<class D >

float Bial::OPF::PDF	(	const Feature< D > &	feature,
		const Adjacency &	adjacency,
		Image< float > &	density,
		float	max_weight
	)

Computes PDF in spatial and spectral domains.

Date: 2012/Nov/26

Parameters

feature	Feature vector.
adjacency	adjacency relation for spacial constraints.
density	Samples density. Used as value map in IFT.
max_weight	maximum arc weight in feature space.

Returns: The bucket size for the IFT queue and PathFunction.

Warning: none.

template<class D >

float Bial::OPF::PDFThread	(	const Feature< D > &	feature,
		const Adjacency &	adjacency,
		Image< float > &	density,
		float	sigma,
		float &	max_dens_diff,
		size_t	thread,
		size_t	total_threads
	)

Computes PDF in spatial and spectral domains.

Date: 2013/Dec/10

Parameters

feature	Feature vector.
adjacency	adjacency relation for spacial constraints.
density	Samples density. Used as value map in IFT.
sigma	Sigma(variance) value for exponential PDF.
max_dens_diff	Maximum density difference found among samples.
thread	Thread number.
total_threads	Number of threads.

Returns: The bucket size for the IFT queue and PathFunction.

Warning: none.

template<class D >

size_t Bial::OPF::SpacialClustering	(	const Feature< D > &	feature,
		Image< int > &	label,
		const Adjacency &	adjacency,
		float	intensity_fraction
	)

Computes OPF clustering based on the feature and image space, using complete graph.

Date: 2013/Nov/26

Parameters

feature	Feature vector.
label	label map.
adjacency	adjacency relation for spacial constraints.
intensity_fraction	fraction from 0.0 to 1.0 of the maximum intensity set for adjacent pixels.

Returns: Number of clusters.

Warning: Feature and init label must have compatible dimensions.

template<class D >

size_t Bial::OPF::SpectralClustering	(	Feature< D > &	feature,
		float	scale_min,
		float	scale_max
	)

Computes OPF clustering based only on the feature space, using sampling and a graph.

Date: 2013/Jul/05

Parameters

feature	Feature vector.
scale_min,scale_max	minimum and maximum scale. From 0.0 to 1.0.
graph	the graph used to compute OPF. Default: KnnGraph for function with no parameter.

Returns: Number of clusters.

Warning: scale_min must be smaller than of equal to scale_min.

template<class D , class GRAPH_ADJACENCY >

size_t Bial::OPF::SpectralClustering	(	Feature< D > &	feature,
		float	scale_min,
		float	scale_max,
		Graph< GRAPH_ADJACENCY > &	graph
	)

template<class D >

size_t Bial::OPF::SpectralClustering	(	Feature< D > &	feature,
		size_t	clusters
	)

Computes OPF clustering based only on the feature space, using sampling and k-nn arcs.

Date: 2014/Apr/30

Parameters

feature	Feature vector.
clusters	Number of expected clusters.
graph	the graph used to compute OPF.

Returns: Number of clusters.

Warning: none.

template<class D , class GRAPH_ADJACENCY >

size_t Bial::OPF::SpectralClustering	(	Feature< D > &	feature,
		size_t	clusters,
		Graph< GRAPH_ADJACENCY > &	graph
	)

Functions

Function Documentation