org.openimaj.image.analysis.algorithm.histogram

Class InterpolatedBinnedWindowedExtractor

java.lang.Object

org.openimaj.image.analysis.algorithm.histogram.BinnedWindowedExtractor

org.openimaj.image.analysis.algorithm.histogram.InterpolatedBinnedWindowedExtractor

All Implemented Interfaces:

ImageAnalyser<FImage>, WindowedHistogramExtractor

public class InterpolatedBinnedWindowedExtractor
extends BinnedWindowedExtractor

The InterpolatedBinnedWindowedExtractor is an extension to a BinnedWindowedExtractor that performs soft assignment to the histogram bins through linear interpolation. If a point being histogrammed lies directly between two bins, half of its weight will be added to both bins. If a point is directly in the centre of a bin, then its full weight will be added to both bins. All other cases use linear interpolation to distribute the weight across the two nearest bins.

Cyclic histograms are also supported (i.e. for angles).

For non cyclic histograms, the bin centres are at min + binWidth/2 + n*binWidth for n=0... Any point less than binWidth/2 from the end of a non-cyclic histogram counts fully to the respective end bin.

For cyclic histograms, the bin centres are at min + n*binWidth for n=0...

Author:

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

Field Summary

Fields inherited from class org.openimaj.image.analysis.algorithm.histogram.BinnedWindowedExtractor

binMap, max, min, nbins

Constructor Summary

Constructors

Constructor and Description
InterpolatedBinnedWindowedExtractor(int nbins) Construct with the given number of bins.
InterpolatedBinnedWindowedExtractor(int nbins, boolean wrap) Construct with the given number of bins.
InterpolatedBinnedWindowedExtractor(int nbins, float min, float max) Construct with the given number of bins, and range.
InterpolatedBinnedWindowedExtractor(int nbins, float min, float max, boolean wrap) Construct with the given number of bins, and range.

Method Summary

Methods

Modifier and Type	Method and Description
void	analyseImage(FImage image) Computes the bin-map for this image.
Histogram	computeHistogram(int x, int y, FImage extWeights, FImage windowWeights) Compute the histogram for the given window.
Histogram	computeHistogram(int x, int y, int w, int h) Compute the histogram for the given window.
Histogram	computeHistogram(int x, int y, int w, int h, FImage extWeights) Compute the histogram for the given window.
float[][]	getWeightsMap() Get the weights map

Methods inherited from class org.openimaj.image.analysis.algorithm.histogram.BinnedWindowedExtractor

computeHistogram, computeHistogram, computeHistogram, computeHistogram, getBinMap, getMax, getMin, getNumBins, setMax, setMin, setNbins

Methods inherited from class java.lang.Object

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

Constructor Detail

InterpolatedBinnedWindowedExtractor

public InterpolatedBinnedWindowedExtractor(int nbins)

Construct with the given number of bins. The histogram is not cyclic. The minimum expected value is assumed to be 0 and the maximum 1.

Parameters:

nbins - number of bins

InterpolatedBinnedWindowedExtractor

public InterpolatedBinnedWindowedExtractor(int nbins,
                                   boolean wrap)

Construct with the given number of bins. The histogram is optionally cyclic. The minimum expected value is assumed to be 0 and the maximum 1.

Parameters:

nbins - number of bins

wrap - true if the histogram is cyclic; false otherwise

InterpolatedBinnedWindowedExtractor

public InterpolatedBinnedWindowedExtractor(int nbins,
                                   float min,
                                   float max)

Construct with the given number of bins, and range. The histogram is not cyclic.

Parameters:

nbins - number of bins

min - minimum expected value

max - maximum expected value

InterpolatedBinnedWindowedExtractor

public InterpolatedBinnedWindowedExtractor(int nbins,
                                   float min,
                                   float max,
                                   boolean wrap)

Construct with the given number of bins, and range. The histogram is optionally cyclic.

Parameters:

nbins - number of bins

min - minimum expected value

max - maximum expected value

wrap - true if the histogram is cyclic; false otherwise

Method Detail

analyseImage

public void analyseImage(FImage image)

Computes the bin-map for this image.

Specified by:

analyseImage in interface ImageAnalyser<FImage>

Overrides:

analyseImage in class BinnedWindowedExtractor

Parameters:

image - The image to process in place.

computeHistogram

public Histogram computeHistogram(int x,
                         int y,
                         int w,
                         int h)

Description copied from interface: WindowedHistogramExtractor

Compute the histogram for the given window.

Specified by:

computeHistogram in interface WindowedHistogramExtractor

Overrides:

computeHistogram in class BinnedWindowedExtractor

Parameters:

x - The x-coordinate of the top-left of the window

y - The y-coordinate of the top-left of the window

w - The width of the window

h - The height of the window

Returns:

the histogram in the window of the last analysed image

computeHistogram

public Histogram computeHistogram(int x,
                         int y,
                         int w,
                         int h,
                         FImage extWeights)

Description copied from class: BinnedWindowedExtractor

Compute the histogram for the given window. The weight for each bin is taken from the given weights image.

Overrides:

computeHistogram in class BinnedWindowedExtractor

Parameters:

x - The x-coordinate of the top-left of the window

y - The y-coordinate of the top-left of the window

w - The width of the window

h - The height of the window

extWeights - the weights image. Must be the same size as the analysed image.

Returns:

the histogram in the window of the last analysed image

computeHistogram

public Histogram computeHistogram(int x,
                         int y,
                         FImage extWeights,
                         FImage windowWeights)

Description copied from class: BinnedWindowedExtractor

Compute the histogram for the given window. The weight for each bin is taken from the given weights image, and is multiplied by the corresponding weight in the window image before accumulation. The size of the window is taken from the window weights image.

This method primarily allows you to compute a spatially weighted histogram. For example, the window weights image could be a 2D Gaussian, and thus the histogram would apply more weight on to the centre pixels.

Overrides:

computeHistogram in class BinnedWindowedExtractor

Parameters:

x - The x-coordinate of the top-left of the window

y - The y-coordinate of the top-left of the window

extWeights - The weights image. Must be the same size as the analysed image.

windowWeights - The weights for each pixel in the window.

Returns:

the histogram in the window of the last analysed image

getWeightsMap

public float[][] getWeightsMap()

Get the weights map

Returns:

the weights map