Package com.crawljax.stateabstractions.visual

Class PerceptualImageDifferencing

java.lang.Object

com.crawljax.stateabstractions.visual.PerceptualImageDifferencing

public class PerceptualImageDifferencing
extends Object

Adapted from Perceptual Image Difference Utility.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	PerceptualImageDifferencing.Builder	Builds parameter list for the PerceptualDiff constructor.
protected class	PerceptualImageDifferencing.Comparison	ForkJoin idioms adapted from sumOfSquares sample.
private class	PerceptualImageDifferencing.PyramidTask	Converts color values and constructs Laplacian pyramids.

Field Summary

Fields

Modifier and Type	Field	Description
private int	adaptationLevel
private static int	ALPHA_MASK
static int	COLOR_FAIL
static int	COLOR_PASS
private double	colorFactor
private static double	EPSILON
private boolean	failFast
private static boolean	FAST_POW
private double	fieldOfView
private double	gamma
private static double	KAPPA
private static float[]	KERNEL	Filter kernel for Laplacian convolution.
private double	luminance
private boolean	luminanceOnly
private double[]	lut
private static int	MAX_PYR_LEVELS
private double	numOneDegreePixels
private int	thresholdPixels
private static double	XW
private static double	YW
private static double	ZW

Constructor Summary

Constructors

Constructor	Description
PerceptualImageDifferencing(double fieldOfView, int thresholdPixels, boolean failFast, double gamma, double luminance, boolean luminanceOnly, double colorFactor)	Creates PerceptualDiff instance with specified parameters.

Method Summary

Modifier and Type	Method	Description
void	compare(ForkJoinPool pool, BufferedImage imgA, BufferedImage imgB, BufferedImage imgDiff)	Compares images using Yee's method.
protected static void	construct(float[][] levels, int width, int height)	Constructs the Laplacian pyramid by successively copying earlier levels and blurring them.
protected void	convert(int[] rgb, float[] a, float[] b, float[] lum)	Converts RGB to AB and luminance.
private static void	convolveAndTranspose(float[] src, float[] dst, int width, int height)	Convolves image src with 1D filter kernel and stores it transposed in dst.
private static double	csf(double cpd, double lum)	Computes the contrast sensitivity function (Barten SPIE 1989) given the cycles per degree (cpd) and luminance (lum).
void	dump()	Prints some parameters to the log.
private static double	fastpow(double a, double b)	Returns the first argument raised to the power of the second argument.
private static double	mask(double contrast)	Visual Masking Function from Daly 1993
private static double	pow(double a, double b)
private static double	tvi(double adaptationLuminance)	Given the adaptation luminance, computes the threshold of visibility in cd per m^2.

Methods inherited from class java.lang.Object

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

Field Detail

COLOR_PASS

public static final int COLOR_PASS

See Also:

Constant Field Values

COLOR_FAIL

public static final int COLOR_FAIL

See Also:

Constant Field Values

MAX_PYR_LEVELS

private static final int MAX_PYR_LEVELS

See Also:

Constant Field Values

ALPHA_MASK

private static final int ALPHA_MASK

See Also:

Constant Field Values

XW

private static final double XW

See Also:

Constant Field Values

YW

private static final double YW

See Also:

Constant Field Values

ZW

private static final double ZW

See Also:

Constant Field Values

EPSILON

private static final double EPSILON

See Also:

Constant Field Values

KAPPA

private static final double KAPPA

See Also:

Constant Field Values

KERNEL

private static final float[] KERNEL

Filter kernel for Laplacian convolution.

FAST_POW

private static final boolean FAST_POW

See Also:

Constant Field Values

colorFactor

private final double colorFactor

fieldOfView

private final double fieldOfView

gamma

private final double gamma

luminance

private final double luminance

luminanceOnly

private final boolean luminanceOnly

thresholdPixels

private final int thresholdPixels

failFast

private final boolean failFast

numOneDegreePixels

private final double numOneDegreePixels

adaptationLevel

private final int adaptationLevel

lut

private final double[] lut

Constructor Detail

PerceptualImageDifferencing

public PerceptualImageDifferencing(double fieldOfView,
                                   int thresholdPixels,
                                   boolean failFast,
                                   double gamma,
                                   double luminance,
                                   boolean luminanceOnly,
                                   double colorFactor)

Creates PerceptualDiff instance with specified parameters.

Parameters:

fieldOfView - field of view in degrees

thresholdPixels - failure threshold

failFast - whether to fail immediately when threshold is reached

gamma - value to convert rgb into linear space

luminance - white luminance

luminanceOnly - whether to ignore chroma in comparison

colorFactor - how much of color to use

Method Detail

construct

protected static void construct(float[][] levels,
                                int width,
                                int height)

Constructs the Laplacian pyramid by successively copying earlier levels and blurring them.

convolveAndTranspose

private static void convolveAndTranspose(float[] src,
                                         float[] dst,
                                         int width,
                                         int height)

Convolves image src with 1D filter kernel and stores it transposed in dst.

Adapted from Jerry Huxtable's Gaussian Filter

tvi

private static double tvi(double adaptationLuminance)

Given the adaptation luminance, computes the threshold of visibility in cd per m^2.

TVI means Threshold vs Intensity function.

This version comes from Ward Larson Siggraph 1997

csf

private static double csf(double cpd,
                          double lum)

Computes the contrast sensitivity function (Barten SPIE 1989) given the cycles per degree (cpd) and luminance (lum).

mask

private static double mask(double contrast)

Visual Masking Function from Daly 1993

pow

private static double pow(double a,
                          double b)

fastpow

private static double fastpow(double a,
                              double b)

Returns the first argument raised to the power of the second argument.

This approximate implementation from martin.ankerl.com is several times faster than Math.pow.

For an implementation with adjustable accuracy, see article by Harrison Ainsworth: Fast pow() With Adjustable Accuracy.

Parameters:

a - the base

b - the exponent

Returns:

the value a^b

dump

public void dump()

Prints some parameters to the log.

compare

public void compare(ForkJoinPool pool,
                    BufferedImage imgA,
                    BufferedImage imgB,
                    BufferedImage imgDiff)

Compares images using Yee's method.

References: A Perceptual Metric for Production Testing, Hector Yee, Journal of Graphics Tools 2004.

Parameters:

pool - fork-join pool for task execution

imgA - first of two images to compare

imgB - second of two images to compare

imgDiff - accumulates differences (optional)

convert

protected void convert(int[] rgb,
                       float[] a,
                       float[] b,
                       float[] lum)

Converts RGB to AB and luminance.

Parameters:

rgb - pixel values

a - A

b - B

lum - Y * luminance