org.openimaj.image.model.asm

Class MultiResolutionActiveShapeModel<I extends org.openimaj.image.Image<?,I> & org.openimaj.image.processor.SinglebandImageProcessor.Processable<Float,org.openimaj.image.FImage,I>>

java.lang.Object

org.openimaj.image.model.asm.MultiResolutionActiveShapeModel<I>

Type Parameters:

I - Concrete type of Image

@Reference(author={"Cootes, T. F.","Taylor, C. J."},title="Statistical Models of Appearance for Computer Vision",type=Unpublished,month="October",year="2001",url="http://isbe.man.ac.uk/~bim/Models/app_model.ps.gz") @Reference(type=Inproceedings,author={"Cootes, T F","Taylor, C J","Lanitis, A"},title="Active shape models: Evaluation of a multi-resolution method for improving image search",year="1994",booktitle="Proc British Machine Vision Conference",pages={"327","","336"},url="http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.141.4937&rep=rep1&type=pdf",editor="Hancock, E",publisher="BMVA Press",volume="1")
public class MultiResolutionActiveShapeModel<I extends org.openimaj.image.Image<?,I> & org.openimaj.image.processor.SinglebandImageProcessor.Processable<Float,org.openimaj.image.FImage,I>>
extends Object

Implementation of a basic Multi-resolution Active Shape Model. The implementation allows different types of landmark appearance models and can work with both colour and greylevel images.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
MultiResolutionActiveShapeModel(ActiveShapeModel<I>[] asms) Construct a MultiResolutionActiveShapeModel from the stack of provided ActiveShapeModels.

Method Summary

Modifier and Type	Method and Description
ActiveShapeModel.IterationResult	fit(I initialImage, PointList initialShape) Perform multi-resolution fitting of the initial shape to the initial image.
org.openimaj.math.geometry.shape.PointDistributionModel	getPDM()
static <I extends org.openimaj.image.Image<?,I> & org.openimaj.image.processor.SinglebandImageProcessor.Processable<Float,org.openimaj.image.FImage,I>> MultiResolutionActiveShapeModel<I>	trainModel(int numLevels, org.openimaj.math.matrix.algorithm.pca.PrincipalComponentAnalysis.ComponentSelector selector, List<org.openimaj.util.pair.IndependentPair<PointList,I>> data, org.openimaj.math.geometry.shape.PointDistributionModel.Constraint constraint, LandmarkModelFactory<I> factory) Train a new MultiResolutionActiveShapeModel from the given data.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

MultiResolutionActiveShapeModel

public MultiResolutionActiveShapeModel(ActiveShapeModel<I>[] asms)

Construct a MultiResolutionActiveShapeModel from the stack of provided ActiveShapeModels. The ASMs should be arranged in order of decreasing resolution.

Parameters:

asms -

Method Detail

trainModel

public static <I extends org.openimaj.image.Image<?,I> & org.openimaj.image.processor.SinglebandImageProcessor.Processable<Float,org.openimaj.image.FImage,I>> MultiResolutionActiveShapeModel<I> trainModel(int numLevels,
                                                                                                                                                                                                             org.openimaj.math.matrix.algorithm.pca.PrincipalComponentAnalysis.ComponentSelector selector,
                                                                                                                                                                                                             List<org.openimaj.util.pair.IndependentPair<PointList,I>> data,
                                                                                                                                                                                                             org.openimaj.math.geometry.shape.PointDistributionModel.Constraint constraint,
                                                                                                                                                                                                             LandmarkModelFactory<I> factory)

Train a new MultiResolutionActiveShapeModel from the given data.

Type Parameters:

I - Concrete type of Image

Parameters:

numLevels - number of levels in the pyramid (scales)

selector - a PrincipalComponentAnalysis.ComponentSelector for choosing significant EVs

data - annotated images for training

constraint - a PointDistributionModel.Constraint for constraining plausible shapes from the PointDistributionModel.

factory - a LandmarkModelFactory for producing models of local appearance around the landmarks.

Returns:

a newly trained MultiResolutionActiveShapeModel.

fit

public ActiveShapeModel.IterationResult fit(I initialImage,
                                            PointList initialShape)

Perform multi-resolution fitting of the initial shape to the initial image.

Parameters:

initialImage - the initial shape.

initialShape - the initial image.

Returns:

the fitted model parameters.

getPDM

public org.openimaj.math.geometry.shape.PointDistributionModel getPDM()

Returns:

the PointDistributionModel