org.openimaj.image.model.asm

Class ActiveShapeModel<I extends Image<?,I>>

java.lang.Object

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

Type Parameters:

I - Concrete type of Image

@References(references={@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T. F.","Taylor, C. J."},title="Statistical Models of Appearance for Computer Vision",type=Unpublished,month="October",year="2001",url="http://isbe.man.ac.uk/~bim/Models/app_model.ps.gz"),@Reference(type=Inproceedings,author={"T. F. Cootes","C. J. Taylor"},title="Active Shape Models",year="1992",booktitle="in Proceedings of the British Machine Vision Conference")})
public class ActiveShapeModel<I extends Image<?,I>>
extends Object

Implementation of a basic 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)

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ActiveShapeModel.IterationResult Class to hold the response of a single iteration of model fitting.

Constructor Summary

Constructors

Constructor and Description
ActiveShapeModel(PointDistributionModel pdm, LandmarkModel<I>[] landmarkModels) Construct an ActiveShapeModel from a pre-trained PointDistributionModel and set of LandmarkModels.

Method Summary

Methods

Modifier and Type	Method and Description
ActiveShapeModel.IterationResult	fit(I image, PointList initialShape) Iteratively apply performIteration(Image, PointList) until the maximum number of iterations is exceeded, or the number of points that moved less than 0.5 of their maximum distance in an iteration is less than the target inlier percentage.
double	getInlierPercentage()
LandmarkModel<I>[]	getLandmarkModels()
int	getMaxIterations()
PointDistributionModel	getPDM()
ActiveShapeModel.IterationResult	performIteration(I image, PointList currentShape) Perform a single iteration of model fitting.
void	setInlierPercentage(double inlierPercentage) Set the target percentage of the number of points that move less than 0.5 of their total possible distance within an iteration to stop fitting.
void	setMaxIterations(int maxIter) Set the maximum allowed number of iterations in fitting the model
static <I extends Image<?,I>> ActiveShapeModel<I>	trainModel(PrincipalComponentAnalysis.ComponentSelector selector, List<IndependentPair<PointList,I>> data, PointDistributionModel.Constraint constraint, LandmarkModelFactory<I> factory) Train a new ActiveShapeModel using the given data and parameters.

Methods inherited from class java.lang.Object

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

Constructor Detail

ActiveShapeModel

public ActiveShapeModel(PointDistributionModel pdm,
                LandmarkModel<I>[] landmarkModels)

Construct an ActiveShapeModel from a pre-trained PointDistributionModel and set of LandmarkModels.

Parameters:

pdm - the PointDistributionModel.

landmarkModels - the LandmarkModels.

Method Detail

trainModel

public static <I extends Image<?,I>> ActiveShapeModel<I> trainModel(PrincipalComponentAnalysis.ComponentSelector selector,
                                                    List<IndependentPair<PointList,I>> data,
                                                    PointDistributionModel.Constraint constraint,
                                                    LandmarkModelFactory<I> factory)

Train a new ActiveShapeModel using the given data and parameters.

Type Parameters:

I - The concrete image type.

Parameters:

selector - the selector for choosing the number of principal components / modes of the model.

data - the data to train the model from

constraint - the constraint to apply to restrict the model to plausible shapes.

factory - the LandmarkModelFactory for learning local appearance models

Returns:

a newly trained ActiveShapeModel.

performIteration

public ActiveShapeModel.IterationResult performIteration(I image,
                                                PointList currentShape)

Perform a single iteration of model fitting.

Parameters:

image - the image to fit to

currentShape - the starting shape in image coordinates

Returns:

the updated shape and parameters

fit

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

Iteratively apply performIteration(Image, PointList) until the maximum number of iterations is exceeded, or the number of points that moved less than 0.5 of their maximum distance in an iteration is less than the target inlier percentage.

Parameters:

image - the image to fit the shape to

initialShape - the initial shape in image coordinates

Returns:

the fitted shape and parameters

See Also:

setInlierPercentage(double), setMaxIterations(int)

getMaxIterations

public int getMaxIterations()

Returns:

the maxIter

setMaxIterations

public void setMaxIterations(int maxIter)

Set the maximum allowed number of iterations in fitting the model

Parameters:

maxIter - the maxIter to set

getInlierPercentage

public double getInlierPercentage()

Returns:

the inlierPercentage

setInlierPercentage

public void setInlierPercentage(double inlierPercentage)

Set the target percentage of the number of points that move less than 0.5 of their total possible distance within an iteration to stop fitting.

Parameters:

inlierPercentage - the inlierPercentage to set

getPDM

public PointDistributionModel getPDM()

Returns:

the learnt PointDistributionModel

getLandmarkModels

public LandmarkModel<I>[] getLandmarkModels()

Returns:

the local landmark appearance models; one for each point in the shape.