Set of functions for efficient connectivity analysis of binary or thresholded images. More...

Detailed Description

Set of functions for efficient connectivity analysis of binary or thresholded images.

This namespace provides optimized implementations for identifying and labeling connected components in 2D and 3D images. Both CPU and GPU implementations are available for both 2D and 3D images.

Enumerations
enum	Connectivity { CONNECT4_2D = 4 , CONNECT8_2D = 8 , CONNECT6_3D = 6 , CONNECT18_3D = 18 , CONNECT26_3D = 26 }
	Connectivity types for 2D and 3D connected component analysis. More...

Functions
template<typename T, typename S>
int	computeConnectedComponents (const TypedImage< T > inImage, TypedImage< S > labelImage, Connectivity connectivity, int minComponentSize=-1, int maxNumberComponents=-1, T threshold=T(1), int maxComponentSize=-1, std::vector< size_t > *componentSizes=nullptr)
	Compute connected components of the input image using a CPU-based implementation.
template<typename S>
int	computeConnectedComponents (GlImage inImage, TypedImage< S > labelImage, Connectivity connectivity, int minComponentSize=-1, int maxNumberComponents=-1, double threshold=1.0, int maxComponentSize=-1, std::vector< size_t > *componentSizes=nullptr)
	Compute connected components of the input image using a GPU-accelerated implementation.
template<typename T, typename S>
int	hysteresisThresholding (TypedImage< T > inImage, TypedImage< S > labelImage, double lowThreshold, double highThreshold, Connectivity connectivity, int minComponentSize=-1, int maxNumberComponents=-1)

Enumeration Type Documentation

◆ Connectivity

enum ImFusion::ConnectivityAnalysis::Connectivity

Connectivity types for 2D and 3D connected component analysis.

Defines the neighborhood structure used to determine if pixels/voxels belong to the same component:

2D: 4-connectivity (face neighbors) or 8-connectivity (face + diagonal neighbors)
3D: 6-connectivity (face neighbors), 18-connectivity (face + edge neighbors), or 26-connectivity (face + edge + corner neighbors)

Enumerator
CONNECT4_2D	2D connectivity using 4 face neighbors
CONNECT8_2D	2D connectivity using 8 face and diagonal neighbors
CONNECT6_3D	3D connectivity using 6 face neighbors
CONNECT18_3D	3D connectivity using 18 face and edge neighbors
CONNECT26_3D	3D connectivity using 26 face, edge, and corner neighbors

Function Documentation

◆ computeConnectedComponents() [1/2]

template<typename T, typename S>

int ImFusion::ConnectivityAnalysis::computeConnectedComponents	(	const TypedImage< T > *	inImage,
		TypedImage< S > *	labelImage,
		Connectivity	connectivity,
		int	minComponentSize = -1,
		int	maxNumberComponents = -1,
		T	threshold = T(1),
		int	maxComponentSize = -1,
		std::vector< size_t > *	componentSizes = nullptr )

Compute connected components of the input image using a CPU-based implementation.

Identifies and labels all connected components in the input image using an optimized union-find algorithm. All pixels with a value greater than or equal to the specified threshold are considered foreground and processed. Components can be filtered by size and total count. The resulting label image assigns each pixel a unique component ID (starting from 1), with larger components receiving lower IDs. Background pixels are labeled as 0.

Parameters

inImage	Input image to analyze, will not be modified
labelImage	Output image to store component labels; must have the same dimensions as the input image
connectivity	Type of connectivity to use (4 or 8 for 2D images, 6, 18, or 26 for 3D images)
minComponentSize	Minimum component size in pixels/voxels; components smaller than this will be removed (-1 = no filtering)
maxNumberComponents	Maximum number of components to keep, sorted by size (-1 = keep all)
threshold	Threshold value; pixels >= threshold are considered foreground
maxComponentSize	Maximum component size in pixels/voxels; components larger than this will be removed (-1 = no filtering)
componentSizes	Optional output vector to receive component sizes; index i contains the size of component label i+1

Returns: Number of components found after filtering, or -1 on error

◆ computeConnectedComponents() [2/2]

template<typename S>

int ImFusion::ConnectivityAnalysis::computeConnectedComponents	(	GlImage *	inImage,
		TypedImage< S > *	labelImage,
		Connectivity	connectivity,
		int	minComponentSize = -1,
		int	maxNumberComponents = -1,
		double	threshold = 1.0,
		int	maxComponentSize = -1,
		std::vector< size_t > *	componentSizes = nullptr )

Compute connected components of the input image using a GPU-accelerated implementation.

Identifies and labels all connected components in the input image using GPU compute shaders. All pixels with a value greater than or equal to the specified threshold are considered foreground and processed. Components can be filtered by size and total count. The resulting label image assigns each pixel a unique component ID (starting from 1), with larger components receiving lower IDs. Background pixels are labeled as 0. Note: The underlying algorithm is inherently hard to parallelize, so the GPU implementation might be slower than the CPU implementation for small images.