org.openimaj.image.analysis.pyramid.gaussian

Class GaussianPyramidOptions<IMAGE extends Image<?,IMAGE> & SinglebandImageProcessor.Processable<Float,FImage,IMAGE>>

java.lang.Object

org.openimaj.image.analysis.pyramid.PyramidOptions<GaussianOctave<IMAGE>,IMAGE>

org.openimaj.image.analysis.pyramid.gaussian.GaussianPyramidOptions<IMAGE>

Type Parameters:

IMAGE - type of underlying image.

Direct Known Subclasses:

DoGSIFTEngineOptions

public class GaussianPyramidOptions<IMAGE extends Image<?,IMAGE> & SinglebandImageProcessor.Processable<Float,FImage,IMAGE>>
extends PyramidOptions<GaussianOctave<IMAGE>,IMAGE>

Options for constructing a Gaussian pyramid in the style of Lowe's SIFT paper.

Author:

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

Field Summary

Fields

Modifier and Type	Field and Description
protected int	borderPixels Number of pixels of border for processors to ignore.
protected boolean	doubleInitialImage Should the starting image of the pyramid be stretched to twice its size?
protected int	extraScaleSteps The number of extra scale steps taken beyond scales.
protected float	initialSigma Assumed initial scale of the first image in each octave.
protected int	scales The number of scales in this octave minus extraScaleSteps.

Fields inherited from class org.openimaj.image.analysis.pyramid.PyramidOptions

keepOctaves, octaveProcessor, pyramidProcessor

Constructor Summary

Constructors

Constructor and Description
GaussianPyramidOptions() Default constructor.
GaussianPyramidOptions(GaussianPyramidOptions<?> options) Construct the pyramid options by copying the non-processor options from the given options object.

Method Summary

Methods

Modifier and Type	Method and Description
SinglebandImageProcessor<Float,FImage>	createGaussianBlur(float sigma) Create a SinglebandImageProcessor that performs a Gaussian blurring with a standard deviation given by sigma.
int	getBorderPixels() Get the number of pixels used for a border that processors shouldn't touch.
int	getExtraScaleSteps() Get the number of extra scale steps taken beyond scales.
float	getInitialSigma() Get the assumed initial scale of the first image in each octave.
int	getScales() Get the number of scales in this octave minus extraScaleSteps.
boolean	isDoubleInitialImage() Should the starting image of the pyramid be stretched to twice its size?
void	setBorderPixels(int borderPixels) Set the number of pixels used for a border that processors shouldn't touch.
void	setDoubleInitialImage(boolean doubleInitialImage) Set whether starting image of the pyramid be stretched to twice its size?
void	setExtraScaleSteps(int extraScaleSteps) Set the number of extra scale steps taken beyond scales.
void	setInitialSigma(float initialSigma) Set the assumed initial scale of the first image in each octave.
void	setScales(int scales) Set the number of scales in this octave minus extraScaleSteps.

Methods inherited from class org.openimaj.image.analysis.pyramid.PyramidOptions

getOctaveProcessor, getPyramidProcessor, isKeepOctaves, setKeepOctaves, setOctaveProcessor, setPyramidProcessor

Methods inherited from class java.lang.Object

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

Field Detail

borderPixels

protected int borderPixels

Number of pixels of border for processors to ignore. Also used in calculating the minimum image size for the last octave.

doubleInitialImage

protected boolean doubleInitialImage

Should the starting image of the pyramid be stretched to twice its size?

extraScaleSteps

protected int extraScaleSteps

The number of extra scale steps taken beyond scales.

initialSigma

protected float initialSigma

Assumed initial scale of the first image in each octave. For SIFT, Lowe suggested 1.6 (for optimal repeatability; see Lowe's IJCV paper, P.10).

scales

protected int scales

The number of scales in this octave minus extraScaleSteps. Levels are constructed so that level[scales] has twice the sigma of level[0].

Constructor Detail

GaussianPyramidOptions

public GaussianPyramidOptions()

Default constructor.

GaussianPyramidOptions

public GaussianPyramidOptions(GaussianPyramidOptions<?> options)

Construct the pyramid options by copying the non-processor options from the given options object.

Parameters:

options - options to copy from

Method Detail

getBorderPixels

public int getBorderPixels()

Get the number of pixels used for a border that processors shouldn't touch.

Returns:

number of border pixels.

getExtraScaleSteps

public int getExtraScaleSteps()

Get the number of extra scale steps taken beyond scales.

Returns:

the extraScaleSteps

See Also:

getScales()

getInitialSigma

public float getInitialSigma()

Get the assumed initial scale of the first image in each octave.

Returns:

the initialSigma

getScales

public int getScales()

Get the number of scales in this octave minus extraScaleSteps. Levels of each octave are constructed so that level[scales] has twice the sigma of level[0].

Returns:

the scales

isDoubleInitialImage

public boolean isDoubleInitialImage()

Should the starting image of the pyramid be stretched to twice its size?

Returns:

the doubleInitialImage

setBorderPixels

public void setBorderPixels(int borderPixels)

Set the number of pixels used for a border that processors shouldn't touch. Also affects the minimum image size for the last octave, which must be at least 2 + 2*borderPixels.

Parameters:

borderPixels - number of pixels to leave as border

setDoubleInitialImage

public void setDoubleInitialImage(boolean doubleInitialImage)

Set whether starting image of the pyramid be stretched to twice its size?

Parameters:

doubleInitialImage - the doubleInitialImage to set

setExtraScaleSteps

public void setExtraScaleSteps(int extraScaleSteps)

Set the number of extra scale steps taken beyond scales.

Parameters:

extraScaleSteps - the extraScaleSteps to set

See Also:

setScales(int)

setInitialSigma

public void setInitialSigma(float initialSigma)

Set the assumed initial scale of the first image in each octave. For SIFT, Lowe suggested 1.6 (for optimal repeatability; see Lowe's IJCV paper, P.10).

Parameters:

initialSigma - the initialSigma to set

setScales

public void setScales(int scales)

Set the number of scales in this octave minus extraScaleSteps. Levels of each octave are constructed so that level[scales] has twice the sigma of level[0].

Parameters:

scales - the scales to set

createGaussianBlur

public SinglebandImageProcessor<Float,FImage> createGaussianBlur(float sigma)

Create a SinglebandImageProcessor that performs a Gaussian blurring with a standard deviation given by sigma. This method is used by the GaussianOctave and GaussianPyramid to create filters for performing the blurring. By overriding in subclasses, you can control the exact filter implementation (i.e. for speed).

Parameters:

sigma - the gaussian standard deviation

Returns:

the image processor to apply the blur