dlr::computerVision::SegmenterFelzenszwalb< EdgeFunctor > Class Template Reference

This class implements the image segmentation algorithm described [1]. More...

#include <segmenterFelzenszwalb.h>

List of all members.

Public Member Functions
	SegmenterFelzenszwalb (float k=200.0f, float sigma=0.8f, size_t minSegmentSize=20, EdgeFunctor const &edgeFunctor=EdgeFunctor())
template<ImageFormat FORMAT>
std::vector< Edge >	getEdges (const Image< FORMAT > &inImage)
template<ImageFormat FORMAT>
std::vector< Edge >	getEdges4Connected (const Image< FORMAT > &inImage)
template<ImageFormat FORMAT>
std::vector< Edge >	getEdges8Connected (const Image< FORMAT > &inImage)
virtual Array2D< UnsignedInt32 >	getLabelArray ()
virtual Array2D< UnsignedInt32 >	getLabelArray (UnsignedInt32 &numberOfSegments, std::vector< size_t > &segmentSizes)
template<ImageFormat FORMAT>
void	segment (const Image< FORMAT > &inputImage)
template<class ITER >
void	segmentFromEdges (size_t imageRows, size_t imageColumns, ITER edgeBegin, ITER edgeEnd)
Protected Types
typedef privateCode::DisjointSet < float >	Segment
Protected Member Functions
float	getCost (const Segment &C_i, const Segment &C_j)
template<ImageFormat FORMAT>
void	setEdge (Edge &edge, size_t index0, size_t index1, Image< FORMAT > inImage)
void	updateCost (Segment &C_i, float weight)
Protected Attributes
EdgeFunctor	m_edgeFunctor
numeric::Index2D	m_imageSize
float	m_k
size_t	m_minimumSegmentSize
numeric::Array1D< Segment >	m_segmentation
float	m_sigma
size_t	m_smoothSize

---

Detailed Description

template<class EdgeFunctor>
class dlr::computerVision::SegmenterFelzenszwalb< EdgeFunctor >

This class implements the image segmentation algorithm described [1].

Essentially grouping pixels based on local differences so that segmented regions have similar local variances.

Here is an example of how to use this class:

   Image<GRAY8> inputImage0 = readPGM8(getTestImageFileNamePGM0());
   SegmenterFelzenszwalb<EdgeDefaultFunctor> segmenter(
     kappa, sigma, minSegmentSize);
   segmenter.segment(inputImage0);
   Array2D<UnsignedInt32> labelArray = segmenter.getLabelArray();
 @endCode

 If you need more control over the edge weights that are used
 to do the (graph based) segmentation, you can do the
 following:

 @code
   Image<GRAY8> inputImage0 = readPGM8(getTestImageFileNamePGM0());
   SegmenterFelzenszwalb<MyEdgeFunctor> segmenter(
     kappa, sigma, minSegmentSize);
   
   // Generate graph edges for segmentation.
   Image<HSV8> hsvImage = myPreprocessingRoutine(inputImage0);
   std::vector<cv::Edge> edges = segmenter.getEdges(hsvImage);

   // Actually do the segmentation.
   segmenter.segmentFromEdges(hsvImage.rows(), hsvImage.columns(),
                              edges.begin(), edges.end());
   Array2D<UnsignedInt32> labelArray = segmenter.getLabelArray();

[1] Felzenszwalb, P., and Huttenlocher, D., Efficient Graph-Based Image Segmentation, International Journal of Computer Vision, Volume 59, Number 2, September 2004.

Definition at line 87 of file segmenterFelzenszwalb.h.

---

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

• segmenterFelzenszwalb.h