org.openimaj.image.feature.dense.gradient

Class HOG

java.lang.Object

org.openimaj.image.feature.dense.gradient.HOG

All Implemented Interfaces:

ImageAnalyser<FImage>

@Reference(type=Inproceedings,
           author={"Dalal, Navneet","Triggs, Bill"},
           title="Histograms of Oriented Gradients for Human Detection",
           year="2005",
           booktitle="Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR\'05) - Volume 1 - Volume 01",
           pages={"886","","893"},
           url="http://dx.doi.org/10.1109/CVPR.2005.177",
           publisher="IEEE Computer Society",
           series="CVPR \'05",
           customData={"isbn","0-7695-2372-2","numpages","8","doi","10.1109/CVPR.2005.177","acmid","1069007","address","Washington, DC, USA"})
public class HOG
extends Object
implements ImageAnalyser<FImage>

Implementation of an extractor for the Histogram of Oriented Gradients (HOG) feature for object detection. This implementation allows any kind of spatial layout to be used through different implementations of SpatialBinningStrategys. HOG features can be efficiently extracted for many windows of the image.

The actual work of computing and normalising the descriptor is performed by the SpatialBinningStrategy (i.e. a FixedHOGStrategy or FlexibleHOGStrategy); this class just provides the objects required for efficient histogram computation (namely a GradientOrientationHistogramExtractor) for the image being analysed.

Normally, HOG features are computed using all gradients in the image, but this class makes it possible to only consider gradients along "edges" using the analyseImage(FImage, FImage) method.

Author:

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

Field Summary

Fields

Modifier and Type	Field and Description
protected SpatialBinningStrategy	strategy

Constructor Summary

Constructors

Constructor and Description
HOG(int nbins, boolean histogramInterpolation, FImageGradients.Mode orientationMode, SpatialBinningStrategy strategy) Construct a new HOG with the given number of bins.
HOG(SpatialBinningStrategy strategy) Construct a new HOG with the 9 bins, using histogram interpolation and unsigned gradients.

Method Summary

Modifier and Type	Method and Description
void	analyseImage(FImage image) Analyse an image.
void	analyseImage(FImage image, FImage edges) Analyse the given image, but construct the internal data such that the gradient magnitudes are multiplied by the given edge map before being accumulated.
Histogram	getFeatureVector(Rectangle rectangle) Compute the HOG feature for the given window.

Methods inherited from class java.lang.Object

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

Field Detail

strategy

protected SpatialBinningStrategy strategy

Constructor Detail

HOG

public HOG(SpatialBinningStrategy strategy)

Construct a new HOG with the 9 bins, using histogram interpolation and unsigned gradients. Use the given strategy to extract the actual features.

Parameters:

strategy - the SpatialBinningStrategy to use to produce the features

HOG

public HOG(int nbins,
           boolean histogramInterpolation,
           FImageGradients.Mode orientationMode,
           SpatialBinningStrategy strategy)

Construct a new HOG with the given number of bins. Optionally perform linear interpolation across orientation bins. Histograms can also use either signed or unsigned gradients.

Parameters:

nbins - number of bins

histogramInterpolation - if true cyclic linear interpolation is used to share the magnitude across the two closest bins; if false only the closest bin will be filled.

orientationMode - the range of orientations to extract

strategy - the SpatialBinningStrategy to use to produce the features

Method Detail

analyseImage

public void analyseImage(FImage image)

Description copied from interface: ImageAnalyser

Analyse an image.

Specified by:

analyseImage in interface ImageAnalyser<FImage>

Parameters:

image - The image to process in place.

analyseImage

public void analyseImage(FImage image,
                         FImage edges)

Analyse the given image, but construct the internal data such that the gradient magnitudes are multiplied by the given edge map before being accumulated. This could be used to suppress all magnitudes except those at edges; the resultant extracted histograms would only contain information about edge gradients.

Parameters:

image - the image to analyse

edges - the edge image

getFeatureVector

public Histogram getFeatureVector(Rectangle rectangle)

Compute the HOG feature for the given window.

Parameters:

rectangle - the window

Returns:

the computed HOG feature