boofcv.alg.tracker.meanshift

Class LocalWeightedHistogramRotRect<T extends boofcv.struct.image.ImageBase<T>>

java.lang.Object

boofcv.alg.tracker.meanshift.LocalWeightedHistogramRotRect<T>

public class LocalWeightedHistogramRotRect<T extends boofcv.struct.image.ImageBase<T>>
extends java.lang.Object

Computes a local histogram weighted using a Gaussian function. The weighting function is shaped using a rotated rectangle, where the sigma along each axis is set by the rectangle's width and height. For use with TrackerMeanShiftComaniciu2003. The histogram is stored in a row major format.

Field Summary

Fields

Modifier and Type	Field and Description
protected float	c
protected float[]	histogram
float	imageX
float	imageY
protected float	maxPixelValue
protected float	s
protected int[]	sampleHistIndex
protected java.util.List<georegression.struct.point.Point2D_F32>	samplePts
protected float[]	weights

Constructor Summary

Constructors

Constructor and Description
LocalWeightedHistogramRotRect(int numSamples, double numSigmas, int numHistogramBins, int numBands, float maxPixelValue, boofcv.alg.interpolate.InterpolatePixelMB<T> interpolate) Configures histogram calculation.

Method Summary

Modifier and Type	Method and Description
void	computeHistogram(T image, boofcv.struct.RectangleRotate_F32 region) Computes the histogram inside the specified region.
protected int	computeHistogramBin(float[] value) Given the value of a pixel, compute which bin in the histogram it belongs in
protected void	computeHistogramBorder(T image, boofcv.struct.RectangleRotate_F32 region) Computes the histogram and skips pixels which are outside the image border
protected void	computeHistogramInside(boofcv.struct.RectangleRotate_F32 region) Computes the histogram quickly inside the image
protected void	computeWeights(int numSamples, double numSigmas) compute the weights by convolving 1D gaussian kernel
protected void	createSamplePoints(int numSamples) create the list of points in square coordinates that it will sample.
float[]	getHistogram()
int[]	getSampleHistIndex()
java.util.List<georegression.struct.point.Point2D_F32>	getSamplePts()
protected boolean	isInFastBounds(boofcv.struct.RectangleRotate_F32 region) Checks to see if the region can be sampled using the fast algorithm
protected void	normalizeHistogram()
protected void	squareToImageSample(float x, float y, boofcv.struct.RectangleRotate_F32 region) Converts a point from square coordinates into image coordinates

Methods inherited from class java.lang.Object

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

Field Detail

maxPixelValue

protected float maxPixelValue

c

protected float c

s

protected float s

imageX

public float imageX

imageY

public float imageY

sampleHistIndex

protected int[] sampleHistIndex

weights

protected float[] weights

histogram

protected float[] histogram

samplePts

protected java.util.List<georegression.struct.point.Point2D_F32> samplePts

Constructor Detail

LocalWeightedHistogramRotRect

public LocalWeightedHistogramRotRect(int numSamples,
                                     double numSigmas,
                                     int numHistogramBins,
                                     int numBands,
                                     float maxPixelValue,
                                     boofcv.alg.interpolate.InterpolatePixelMB<T> interpolate)

Configures histogram calculation.

Parameters:

numSamples - Number of points it samples along each axis of the rectangle.

numSigmas - Number of standard deviations away the sides will be from the center. Try 3

numHistogramBins - Number of bins in the histogram

numBands - Number of bands in the input image

maxPixelValue - Maximum value of a pixel across all bands

interpolate - Function used to interpolate the image

Method Detail

computeWeights

protected void computeWeights(int numSamples,
                              double numSigmas)

compute the weights by convolving 1D gaussian kernel

createSamplePoints

protected void createSamplePoints(int numSamples)

create the list of points in square coordinates that it will sample. values will range from -0.5 to 0.5 along each axis.

computeHistogram

public void computeHistogram(T image,
                             boofcv.struct.RectangleRotate_F32 region)

Computes the histogram inside the specified region. Results are returned by calling getHistogram().

Parameters:

image - Input image

region - Region which is to be sampled

computeHistogramInside

protected void computeHistogramInside(boofcv.struct.RectangleRotate_F32 region)

Computes the histogram quickly inside the image

computeHistogramBorder

protected void computeHistogramBorder(T image,
                                      boofcv.struct.RectangleRotate_F32 region)

Computes the histogram and skips pixels which are outside the image border

computeHistogramBin

protected int computeHistogramBin(float[] value)

Given the value of a pixel, compute which bin in the histogram it belongs in

isInFastBounds

protected boolean isInFastBounds(boofcv.struct.RectangleRotate_F32 region)

Checks to see if the region can be sampled using the fast algorithm

normalizeHistogram

protected void normalizeHistogram()

squareToImageSample

protected void squareToImageSample(float x,
                                   float y,
                                   boofcv.struct.RectangleRotate_F32 region)

Converts a point from square coordinates into image coordinates

getHistogram

public float[] getHistogram()

getSampleHistIndex

public int[] getSampleHistIndex()

getSamplePts

public java.util.List<georegression.struct.point.Point2D_F32> getSamplePts()