Main Page Class Hierarchy Alphabetical List Compound List Compound Members

prapi::texture::KernelSegmentator Class Reference

#include <Segmentation.h>

Detailed Description

KernelSegmentator performs supervised segmentation on an matrix.

On each pixel of an input matrix, classifier constructs a sample that is classified using a given classifier. The classification result is placed on an output matrix. The sample is constructed by placing a kernel over the input pixel and calculating a histogram over it. The segmentation problem is presented as a layered matrix. To get a histogram bin index for a pixel, the segmentator collects the corresponding pixels from each layer to a vector and consults a LayerCombiner to modify the local histogram.

Public Methods

KernelSegmentator (Classifier< int > &classifier, Matrix< char > kernel, const MultiFeatureHistogram::LayerCombiner &combiner)

Create a new Segmentator.

Matrix< int > getSegmentation (const util::List< Matrix< int > > &mats)

Assign each pixel its classification.

Static Public Methods

Matrix< char > createCircularKernel (int radius)

Create a circular kernel.

Matrix< char > createSquareKernel (int radius)

Create a square kernel.

double getError (const Matrix< int > &segmentation, const Matrix< int > &groundTruth) throw (InvalidArgumentException&)

Get the fraction of misclassified pixels given a ground truth matrix.

Constructor & Destructor Documentation

prapi::texture::KernelSegmentator::KernelSegmentator ( Classifier< int > & classifier,

Matrix< char > kernel,

const MultiFeatureHistogram::LayerCombiner & combiner

) [inline]

Create a new Segmentator.

Parameters:

classifier the classifier to be used in classifying pixels

length the length of the histograms that are created

kernel a kernel matrix with ones on the pixels that are considered in building up the local histogram.

combiner a LayerCombiner that is used to modify the local histogram

Member Function Documentation

Matrix< char > prapi::texture::KernelSegmentator::createCircularKernel ( int radius ) [static]

Create a circular kernel.
The result is a (radius*2+1)x(radius*2+1) matrix with a "disk" of ones in the center.
Parameters:

radius the radius of the kernel disk

Matrix< char > prapi::texture::KernelSegmentator::createSquareKernel ( int radius ) [static]

Create a square kernel.
The result is a (radius*2+1)x(radius*2+1) matrix with with all entries set to ones.
Parameters:

radius the radius of the kernel

double prapi::texture::KernelSegmentator::getError ( const Matrix< int > & segmentation,

const Matrix< int > & groundTruth

) throw (InvalidArgumentException&) [static]

Get the fraction of misclassified pixels given a ground truth matrix.
The ground truth matrix should contain the correct classifications for the pixels.
Exceptions:

InvalidArgumentException & if the matrices are of different size

Matrix< int > prapi::texture::KernelSegmentator::getSegmentation ( const util::List< Matrix< int > > & mats )

Assign each pixel its classification.

Parameters:

mats matrix layers.

combine how to combine layers

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

Segmentation.h
Segmentation.cc

Public Methods
	KernelSegmentator (Classifier< int > &classifier, Matrix< char > kernel, const MultiFeatureHistogram::LayerCombiner &combiner)
	Create a new Segmentator.
Matrix< int >	getSegmentation (const util::List< Matrix< int > > &mats)
	Assign each pixel its classification.
Static Public Methods
Matrix< char >	createCircularKernel (int radius)
	Create a circular kernel.
Matrix< char >	createSquareKernel (int radius)
	Create a square kernel.
double	getError (const Matrix< int > &segmentation, const Matrix< int > &groundTruth) throw (InvalidArgumentException&)
	Get the fraction of misclassified pixels given a ground truth matrix.