Facemorph.tensor
Class Multilinear

java.lang.Object
  Facemorph.tensor.Multilinear

All Implemented Interfaces:

java.io.Serializable

public class Multilinear
extends java.lang.Object
implements java.io.Serializable

Multilinear model class

See Also:

Serialized Form

Constructor Summary

Multilinear(Tensor T, int dataDim, double maxvar)
Creates a new instance of Multilinear Performs PCA along each of the plane average vectors Replace residuals with best linear fit etc etc

Method Summary

double[][]	convertParams(double[] params) Convert the parameters to multi-linear
static double	dotProduct(double[] v1, double[] v2) Find the dot product between two vectors
double[][]	getALSParams(double[] test) Get Alternating Least Squares (ALS) parameters
double[]	getAverage() Get the average
double[][]	getBestFitEdgeParams(double[] test) Gets the best parameters from the edge (linear) components
double[]	getEdgeComp(int index) Get a component along the edges
static Multilinear	getImageMultilinear(int[] dims, java.util.Vector<java.awt.Image> images, double maxvar) Get a multilinear model from a set of images
static Multilinear	getTemplateMultilinear(int[] dims, java.util.Vector<Template> templates, double maxvar) Get a MultiLinear object from the Templates
int	getTotalParams() Get the total number of parameters
static void	main(java.lang.String[] args) Main method for testing
static int[]	readImagesAndTemplates(java.lang.String listFile, java.util.Vector<Template> templates, java.util.Vector<java.awt.Image> images, java.util.Vector<java.lang.String> temNames, java.util.Vector<java.lang.String> imgNames) Read images and templates from a file
double[]	reconstruct(double[][] params) Reconstruct from parameters
double[]	reconstructEdgeParams(double[][] params) Reconstruct using only the edge parameters
double[]	reconvertParams(double[][] params) Creates a single vector from the edge parameters
static void	testLinear(java.lang.String[] args) Test of linear building method
static FloatImage[]	unvectoriseColourFloatImage(double[] rgb, int width, int height) Unvectorise an image to an array of FloatImage
static java.awt.Image	unvectoriseImage(double[] rgb, int width, int height) Unvectorise an image
static double[]	vectoriseColourFloatImage(FloatImage[] rgb) Vectorise 3 plane RGB image into a 1D vector
static double[]	vectoriseImage(java.awt.Image img) Vectorise an image into a 1D vector

Methods inherited from class java.lang.Object

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

Constructor Detail

Multilinear

public Multilinear(Tensor T,
                   int dataDim,
                   double maxvar)
            throws BigMatException

Creates a new instance of Multilinear Performs PCA along each of the plane average vectors Replace residuals with best linear fit etc etc

Parameters:

T - the input Tensor

dataDim - the data dimension

maxvar - the maximum variance to explain in any dimension

Throws:

BigMatException

Method Detail

getAverage

public double[] getAverage()

Get the average

Returns:

return the average

dotProduct

public static double dotProduct(double[] v1,
                                double[] v2)

Find the dot product between two vectors

Parameters:

v1 - the first vector

v2 - the second vector

Returns:

returns the dot product

getEdgeComp

public double[] getEdgeComp(int index)

Get a component along the edges

Parameters:

index - the index of the component

Returns:

returns the edge component

reconstructEdgeParams

public double[] reconstructEdgeParams(double[][] params)

Reconstruct using only the edge parameters

Parameters:

params - the parameters

Returns:

return the reconstructed vector

getTotalParams

public int getTotalParams()

Get the total number of parameters

Returns:

return the total number of parameters

getBestFitEdgeParams

public double[][] getBestFitEdgeParams(double[] test)

Gets the best parameters from the edge (linear) components

Parameters:

test - the vector to analyse

Returns:

return the linear model parameters

getALSParams

public double[][] getALSParams(double[] test)

Get Alternating Least Squares (ALS) parameters

Parameters:

test - the intput vector

Returns:

returns the ALS estimated parameters

reconstruct

public double[] reconstruct(double[][] params)
                     throws BigMatException

Reconstruct from parameters

Parameters:

params - the parameters

Returns:

return the reconstructed vector

Throws:

BigMatException

reconvertParams

public double[] reconvertParams(double[][] params)

Creates a single vector from the edge parameters

Parameters:

params - the intpu Tensor paraameters

Returns:

the vectorised parameters

convertParams

public double[][] convertParams(double[] params)

Convert the parameters to multi-linear

Parameters:

params - the parameters to convert

Returns:

returns the converted parameters

getTemplateMultilinear

public static Multilinear getTemplateMultilinear(int[] dims,
                                                 java.util.Vector<Template> templates,
                                                 double maxvar)
                                          throws BigMatException

Get a MultiLinear object from the Templates

Parameters:

dims - the number of dimensions

templates - the templates

maxvar - the maximum variation to explain

Returns:

return the constructed MultiLinear model

Throws:

BigMatException

vectoriseImage

public static double[] vectoriseImage(java.awt.Image img)

Vectorise an image into a 1D vector

Parameters:

img - the image to vectorise

Returns:

return the vectorised image

unvectoriseImage

public static java.awt.Image unvectoriseImage(double[] rgb,
                                              int width,
                                              int height)

Unvectorise an image

Parameters:

rgb - the data vector to unvectorise

width - the width of the output image

height - the height of the output image

Returns:

return the unvectorised image

unvectoriseColourFloatImage

public static FloatImage[] unvectoriseColourFloatImage(double[] rgb,
                                                       int width,
                                                       int height)

Unvectorise an image to an array of FloatImage

Parameters:

rgb - the vector

width - the output width

height - the output height

Returns:

return the unvectorised image

vectoriseColourFloatImage

public static double[] vectoriseColourFloatImage(FloatImage[] rgb)

Vectorise 3 plane RGB image into a 1D vector

Parameters:

rgb - the intput image in 3 planes

Returns:

return the vectorised image

getImageMultilinear

public static Multilinear getImageMultilinear(int[] dims,
                                              java.util.Vector<java.awt.Image> images,
                                              double maxvar)
                                       throws BigMatException

Get a multilinear model from a set of images

Parameters:

dims - the size of each dimension

images - the images

maxvar - the maximum variance to explain along each dimension

Returns:

returns the multiline model

Throws:

BigMatException

readImagesAndTemplates

public static int[] readImagesAndTemplates(java.lang.String listFile,
                                           java.util.Vector<Template> templates,
                                           java.util.Vector<java.awt.Image> images,
                                           java.util.Vector<java.lang.String> temNames,
                                           java.util.Vector<java.lang.String> imgNames)

Read images and templates from a file

Parameters:

listFile - the file containing the file names

templates - the vector to put the templates in

images - the array to put the images in

temNames - the output vector of template names

imgNames - the output vector of image names

Returns:

returns the model dimensions

main

public static void main(java.lang.String[] args)
                 throws BigMatException

Main method for testing

Parameters:

args - the program args (currently testLinear)

Throws:

BigMatException

testLinear

public static void testLinear(java.lang.String[] args)
                       throws BigMatException

Test of linear building method

Parameters:

args - args[0] - list file, args[1] - width, args[2] - height, args[3] - mask, args[4] - test list

Throws:

BigMatException