Package dev.brachtendorf.jimagehash.hashAlgorithms.filter

Class Kernel

java.lang.Object

dev.brachtendorf.jimagehash.hashAlgorithms.filter.Kernel

All Implemented Interfaces:

Filter, Serializable

Direct Known Subclasses:

Kernel.GrayScaleFilter, MultiKernel, NonAveragingKernel

public class Kernel
extends Object
implements Filter

Kernel operations are shifting window masks applied to data point of an array to recompute it's value. This class supports 2D kernels which are mainly used on images pixel data. e.g. the bellow mask will calculate the value of the pixel by multiplying it's original value by 5 and subtracting surrounding pixels values.

 [  0 -1  0 
   -1  5 -1 
    0 -1  0 ]
 

In image processing sharpening or blurring is realized by applying kernels to pixel data. TODO support separability for custom kernels TODO kernel indices are swapped and twisted.

Since:

2.0.0

Author:

Kilian

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	Kernel.EdgeHandlingStrategy	The edge handling strategy defines the behaviour when a kernel reaches the edge of an image and does not have enough information to compute it's value.
static class	Kernel.GrayScaleFilter	Kernels whose filter class work with grayscale instead of seperate color channels

Field Summary

Fields

Modifier and Type	Field	Description
protected Kernel.EdgeHandlingStrategy	edgeHandling	How are edged of the images handled
protected double[][]	mask	Kernel mask applied to the pixels

Constructor Summary

Constructors

Modifier	Constructor	Description
	Kernel(double[][] mask)	Construct a non normalized kernel with the given pixel mask and the default edge handling strategy of Kernel.EdgeHandlingStrategy.EXPAND.
	Kernel(double[][] mask, boolean normalize)	Construct a kernel with the given pixel mask and the default edge handling strategy of Kernel.EdgeHandlingStrategy.EXPAND.
	Kernel(double[][] mask, Kernel.EdgeHandlingStrategy edgeHandling)	Construct a non normalized kernel with the given pixel mask.
	Kernel(double[][] mask, Kernel.EdgeHandlingStrategy edgeHandling, boolean normalize)	Construct a kernel with the given pixel mask.
	Kernel(Kernel template)	Create a clone of the supplied kernel
protected	Kernel(Kernel.EdgeHandlingStrategy strat)	Deprecated.

Method Summary

Modifier and Type	Method	Description
double[][]	apply(byte[][] input)	Apply the kernel to the 2d array.
double[][]	apply(double[][] input)	Apply the kernel to the 2d array.
double[][]	apply(int[][] input)	Apply the kernel to the 2d array.
byte[][]	applyByte(byte[][] input)	Apply the kernel to the 2d array with each value casted to a byte value.
byte[][]	applyByte(double[][] input)
int[][]	applyInt(double[][] input)	Apply the kernel to the 2d array with each value casted to a int value.
int[][]	applyInt(int[][] input)	Apply the kernel to the 2d array with each value casted to a int value.
static Kernel	boxFilter(int width, int height, double factor)	A box filter is a filter which applies the same factor to squared region of pixels and can be counted to the blurring filters.
static Kernel	boxFilterNormalized(int width, int height)	A box filter is a filter which applies the same factor to squared region of pixels and can be counted to the blurring filters.
static Kernel	boxFilterNormalizedSep(int width, int height)
protected double	calcValue(byte[][] input, int x, int y)
protected double	calcValue(double[][] input, int x, int y)
protected double	calcValue(int[][] input, int x, int y)
static Kernel	edgeDetectionFilter(int strength)	Deprecated.
static Kernel	embossHorizontontalFilter(int depth)
static Kernel	embossLeftDiagonalFilter(int depth)	1 0 0 0 0 0 0 0 -1
static Kernel	embossleftRightFilter(int depth)	0 0 0 1 0 -1 0 0 0
static Kernel	embossRightDiagonalFilter(int depth)	0 0 1 0 0 0 -1 0 0
boolean	equals(Object obj)
BufferedImage	filter(BufferedImage input)	Apply the filter to the input image and return an altered copy
static Kernel	gaussianFilter(int width, int height, double std)	Creates a gaussian blur.
int	hashCode()
static Kernel	identityFilter()	Return an identity kernel.
String	toString()
static Kernel	zeroFilter()	Return an zero kernel.

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Field Detail

mask

protected double[][] mask

Kernel mask applied to the pixels

edgeHandling

protected Kernel.EdgeHandlingStrategy edgeHandling

How are edged of the images handled

Constructor Detail

Kernel

public Kernel(Kernel template)

Create a clone of the supplied kernel

Parameters:

template - the kernel to clone

Kernel

@Deprecated
protected Kernel(Kernel.EdgeHandlingStrategy strat)

Deprecated.

Empty constructor used by inheriting classes which are not able to provide a mask during first constructor call. The inheriting class promises to provide a mask and all checks by itself

Parameters:

strat - EdgeHandlingStrategy to use

Kernel

public Kernel(double[][] mask)

Construct a non normalized kernel with the given pixel mask and the default edge handling strategy of Kernel.EdgeHandlingStrategy.EXPAND.

Parameters:

mask - used to filter pixel. So far only oddly shaped masks are allowed

Throws:

IllegalArgumentException - if mask's width or height is even

Since:

2.0.0

Kernel

public Kernel(double[][] mask,
              boolean normalize)

Construct a kernel with the given pixel mask and the default edge handling strategy of Kernel.EdgeHandlingStrategy.EXPAND.

Parameters:

mask - used to filter pixel. So far only oddly shaped masks are allowed

normalize - If true the mask is normalized resulting in the sum of the mask being 1. This will preserve the magnitude of the original range. If the mask will be blindly copied without adjustment

Throws:

IllegalArgumentException - if mask's width or height is even

Since:

2.0.0

Kernel

public Kernel(double[][] mask,
              Kernel.EdgeHandlingStrategy edgeHandling)

Construct a non normalized kernel with the given pixel mask.

Parameters:

mask - used to filter pixel. So far only oddly shaped masks are allowed

edgeHandling - the edge handling strategy used at the corners of the image

Throws:

IllegalArgumentException - if mask's width or height is even

Since:

2.0.0

Kernel

public Kernel(double[][] mask,
              Kernel.EdgeHandlingStrategy edgeHandling,
              boolean normalize)

Construct a kernel with the given pixel mask.

Parameters:

mask - used to filter pixel. So far only oddly shaped masks are allowed

edgeHandling - the edge handling strategy used at the corners of the image

normalize - If true the mask is normalized resulting in the sum of the mask being 1. This will preserve the magnitude of the original range. If the mask will be blindly copied without adjustment

Throws:

IllegalArgumentException - if mask's width or height is even

Since:

2.0.0

Method Detail

identityFilter

public static Kernel identityFilter()

Return an identity kernel. This kernel is a 1x1 kernel and copies the original value to the new array

Returns:

a identity kernel

zeroFilter

public static Kernel zeroFilter()

Return an zero kernel. This kernel is a 1x1 kernel and zeroes out all values

Returns:

a zero kernel

boxFilter

public static Kernel boxFilter(int width,
                               int height,
                               double factor)

A box filter is a filter which applies the same factor to squared region of pixels and can be counted to the blurring filters.

        Example factor: 0.3 width: 3 height: 3
 
          0.3  0.3  0.3
          0.3  0.3  0.3
          0.3  0.3  0.3
 

A normalized box filter is available via boxFilterNormalized(int, int).

Parameters:

width - of the kernel has to be odd and positive

height - if the kernel has to be odd and positive

factor - the factor applied to each pixel

Returns:

the box filter kernel

boxFilterNormalizedSep

public static Kernel boxFilterNormalizedSep(int width,
                                            int height)

boxFilterNormalized

public static Kernel boxFilterNormalized(int width,
                                         int height)

A box filter is a filter which applies the same factor to squared region of pixels and can be counted to the blurring filters.

This filter is normalized ensuring the same magnitude of the values.

Parameters:

width - of the kernel has to be odd and positive

height - of the kernel has to be odd and positive

Returns:

the box filter kernel

gaussianFilter

public static Kernel gaussianFilter(int width,
                                    int height,
                                    double std)

Creates a gaussian blur.

A gaussian filter blurs the kernel with decreasing amount depending on the distance to the pixel and produces a more natural blurring effect than a box filter.

The gaussian filter is normalized.

Parameters:

width - of the kernel has to be odd and positive

height - of the kernel has to be odd and positive

std - the standard deviation of the kernel. The higher the stronger the blur effect

Returns:

the gaussian kernel

edgeDetectionFilter

@Deprecated
public static Kernel edgeDetectionFilter(int strength)

Deprecated.

embossHorizontontalFilter

public static Kernel embossHorizontontalFilter(int depth)

Parameters:

depth - width and height. higher values will create a stronger effect

Returns:

a kernel creating a horizontal emboss effect

embossLeftDiagonalFilter

public static Kernel embossLeftDiagonalFilter(int depth)

                 1  0  0 
                 0  0  0 
                 0  0 -1
 

Parameters:

depth - width and height. higher values will create a stronger effect

Returns:

a kernel creating a left diagonal emboss effect

embossRightDiagonalFilter

public static Kernel embossRightDiagonalFilter(int depth)

                 0 0 1 
                 0 0 0 
                -1 0 0
 

Parameters:

depth - width and height. higher values will create a stronger effect

Returns:

a kernel creating a right diagonal emboss effect

embossleftRightFilter

public static Kernel embossleftRightFilter(int depth)

                0  0  0 
                1  0 -1 
                0  0  0
 

Parameters:

depth - width and height. higher values will create a stronger effect

Returns:

a kernel creating a right left emboss effect

apply

public double[][] apply(int[][] input)

Apply the kernel to the 2d array. If the desired output is a int[][] array refer to applyInt(int[][]).

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

applyInt

public int[][] applyInt(int[][] input)

Apply the kernel to the 2d array with each value casted to a int value.

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

applyInt

public int[][] applyInt(double[][] input)

Apply the kernel to the 2d array with each value casted to a int value.

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

apply

public double[][] apply(double[][] input)

Apply the kernel to the 2d array.

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

apply

public double[][] apply(byte[][] input)

Apply the kernel to the 2d array. If the desired output is a byte[][] array refer to applyByte(byte[][]).

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

applyByte

public byte[][] applyByte(byte[][] input)

Apply the kernel to the 2d array with each value casted to a byte value.

Parameters:

input - the input array to apply the kernel on

Returns:

a new array created by the kernel

applyByte

public byte[][] applyByte(double[][] input)

calcValue

protected double calcValue(byte[][] input,
                           int x,
                           int y)

calcValue

protected double calcValue(int[][] input,
                           int x,
                           int y)

calcValue

protected double calcValue(double[][] input,
                           int x,
                           int y)

Parameters:

input - array

x - pixelToLookAt

y - pixelToLookAt

Returns:

convolutedPixel fo this x and y

hashCode

public int hashCode()

Overrides:

hashCode in class Object

equals

public boolean equals(Object obj)

Overrides:

equals in class Object

toString

public String toString()

Overrides:

toString in class Object

filter

public BufferedImage filter(BufferedImage input)

Description copied from interface: Filter

Apply the filter to the input image and return an altered copy

Specified by:

filter in interface Filter

Parameters:

input - the input image to apply the filter on

Returns:

the altered image