Proposals:Refactoring Statistics Framework 2007

ITK: Statistical Framework Refactoring

Summary

Motivation

Recently it has been found that certain characteristics of the Statistical Framework API could be improved in order to provide a more consistent inteface.

In particular, issues have been pointed out regarding the following topics

• Iterators
• Samples versus Subsamples
• Statistical calculator integrated into the pipeline

Tasks

The proposed refactoring will include the following specific tasks

• Review all iterators used in the Statistics framework and add tests for their functionalities until bringing their code coverage to 100%
• Restructure the implementation of the Subsample class, to make sure that
  • It can be used in any place where a sample can be used.
  • Sub-sampling can be nested without breaking the consistency of the statistical operation.

Schedule

The refactoring tasks will be initiated on March 23rd and will be completed before April 15th 2007

Specific Program

Tasks

Build a Class Manifesto

There are 72 header files in the Statistics directory.

Class Manifesto

They are categorized in the following table

Conceptual Class	Number
Measurement Containers	8
Frequency Containers	2
Process Objects	2
Traits	1
Mean shift	3
Adaptors	5
Univariate Distributions	4
Algorithms	3
Calculators	6
Generators	12
Filters	10
Multivariate Density Functions	8
Distance Metrics	2
Components	4
Estimators	2
Total	72

Measurement Containers



• KdTree
• Histogram
• VariableDimensionHistogram
• ListSampleBase
• ListSample
• Sample
• Subsample
• MembershipSample

Frequency Containers

• DenseFrequencyContainer
• SparseFrequencyContainer

Process Objects

• SampleClassifier
• SampleClassifierWithMask

Traits

• MeasurementVectorTraits

MeanShift

• ypersphereKernelMeanShiftModeSeeker
• MeanShiftModeCacheMethod
• MeanShiftModeSeekerBase

Adaptors

• ImageToCooccurrenceListAdaptor
• ImageToListAdaptor
• JointDomainImageToListAdaptor
• PointSetToListAdaptor
• ScalarImageToListAdaptor

Univariate Distributions

• ChiSquareDistribution
• GaussianDistribution
• ProbabilityDistribution
• TDistribution

Algorithms

• SampleAlgorithmBase
• StatisticsAlgorithm
• NeighborhoodSampler

Calculators

• CovarianceCalculator
• GreyLevelCooccurrenceMatrixTextureCoefficientsCalculator
• MeanCalculator
• ScalarImageTextureCalculator
• WeightedCovarianceCalculator
• WeightedMeanCalculator

Generators

• ImageToHistogramGenerator
• ImageToListGenerator
• KdTreeGenerator
• ListSampleToHistogramGenerator
• MaskedScalarImageToGreyLevelCooccurrenceMatrixGenerator
• MembershipSampleGenerator
• NormalVariateGenerator
• RandomVariateGeneratorBase
• ScalarImageToGreyLevelCooccurrenceMatrixGenerator
• ScalarImageToHistogramGenerator
• SelectiveSubsampleGenerator
• WeightedCentroidKdTreeGenerator

Filters

* HistogramToEntropyImageFilter

• HistogramToImageFilter
• HistogramToIntensityImageFilter
• HistogramToLogProbabilityImageFilter
• HistogramToProbabilityImageFilter
• ListSampleToHistogramFilter
• SampleMeanShiftBlurringFilter
• SampleMeanShiftClusteringFilter
• SampleSelectiveMeanShiftBlurringFilter
• SampleToHistogramProjectionFilter

Multivariate Density Functions

* DensityFunction

• DistanceToCentroidMembershipFunction
• GaussianDensityFunction
• GoodnessOfFitFunctionBase
• GoodnessOfFitMixtureModelCostFunction
• LogLikelihoodGoodnessOfFitFunction
• MahalanobisDistanceMembershipFunction
• MembershipFunctionBase

Distance Metrics

• DistanceMetric
• EuclideanDistance

Components

• GaussianGoodnessOfFitComponent
• GaussianMixtureModelComponent
• GoodnessOfFitComponentBase
• MixtureModelComponentBase

Estimators

• ExpectationMaximizationMixtureModelEstimator
• KdTreeBasedKmeansEstimator

Class Diagrams