boofcv.alg.feature.dense

Class DescribeDenseHogAlg<Input extends boofcv.struct.image.ImageBase,Derivative extends boofcv.struct.image.ImageBase>

java.lang.Object

boofcv.alg.feature.dense.DescribeDenseHogAlg<Input,Derivative>

Direct Known Subclasses:

DescribeDenseHogAlg_F32, DescribeDenseHogAlg_PLF32, DescribeDenseHogAlg_PLU8, DescribeDenseHogAlg_U8

public abstract class DescribeDenseHogAlg<Input extends boofcv.struct.image.ImageBase,Derivative extends boofcv.struct.image.ImageBase>
extends java.lang.Object

Implementation of the Histogram of Oriented Gradients (HOG) [1] dense feature descriptor. Several variants are described in the paper. The algorithm used here is the "R-HOG unsigned orientation" variant. The descriptor is computed from a regular grid of cells and an unsigned histogram is computed. Unsigned as in the angle is from 0 to 180 degrees instead of 0 to 360.

This is a (hopefully) faithful implementation to the algorithm described in the paper. The descriptors are computed with the following steps.

1. Compute image gradient using [-1,0,1] kernel
2. Compute histogram for each cell in the image
3. Compute descriptor from blocks of cells
4. Normalize descriptor using SIFT style L2-Hys normalization

Cell Histogram

The image is broken up into a regular grid of "cells". Every cell is square with a width of N pixels, where N is a user specified parameter. Within each cell a histogram is computed. The number of histogram bins B is specified by the user. The histogram is created by computing the Euclidean norm for each pixel's gradient inside the cell and then using bilinear interpolation to put it into the two neighboring bins which best match its angle.

Descriptor Blocks

Square blocks, which are M cells wide, are used to compute each descriptor. The histogram is copied into a TupleDesc_F64 in a row-major fashion and a weight is applied. The weight is specified for each cell based on its distance from the block's center. Gaussian weight is used with a standard deviation of 0.5*block_width. Note that blocks do overlap. See paper for details.

Descriptor Normalization

First L2-normalization is applied to the descriptor. Then min(0.2,desc[i]) is applied to all elements in the descriptor. After which L2-normalization is applied again.

Accessing Results

A list of descriptor and their locations is available. The location refers to the top-left most pixel in the region the descriptor is computed from. These lists are computed in a regular grid with row-major ordering. A request can be made for all descriptors computed from inside a rectangular region.

Multi-Band Images

The gradient is computed for each band individually. The band with the largest magnitude at that specific pixel is used as the gradient for the pixel.

[1] Dalal, Navneet, and Bill Triggs. "Histograms of oriented gradients for human detection." Computer Vision and Pattern Recognition, 2005. CVPR 2005.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	DescribeDenseHogAlg.Cell

Field Summary

Fields

Modifier and Type	Field and Description
protected Derivative	derivX
protected Derivative	derivY
protected float	pixelDX
protected float	pixelDY

Constructor Summary

Constructors

Constructor and Description
DescribeDenseHogAlg(int orientationBins, int widthCell, int widthBlock, int stepBlock, boofcv.struct.image.ImageType<Input> imageType) Configures HOG descriptor computation

Method Summary

Modifier and Type	Method and Description
Derivative	_getDerivX()
Derivative	_getDerivY()
float	_getPixelDX()
float	_getPixelDY()
protected void	computeCellWeights() Computes the value of weights inside of a block
abstract void	computeDerivative(int pixelIndex) Computes and stores the gradient at the specified pixel.
TupleDesc_F64	createDescription()
DescribeDenseHogAlg.Cell	getCell(int row, int col)
int	getCellCols()
int	getCellRows()
org.ddogleg.struct.FastQueue<TupleDesc_F64>	getDescriptions() List of descriptors
void	getDescriptorsInRegion(int pixelX0, int pixelY0, int pixelX1, int pixelY1, java.util.List<TupleDesc_F64> output) Convenience function which returns a list of all the descriptors computed inside the specified region in the image
boofcv.struct.image.ImageType<Input>	getImageType()
org.ddogleg.struct.FastQueue<georegression.struct.point.Point2D_I32>	getLocations() List of locations for each descriptor.
int	getOrientationBins()
int	getRegionWidthPixel() Returns the number of pixel's wide the square region is that a descriptor was computed from
int	getStepBlock()
int	getWidthBlock()
int	getWidthCell()
void	process() Computes the descriptor across the input image
void	setInput(Input input) Specifies input image.
void	setStepBlock(int stepBlock)
void	setWidthCell(int widthCell)

Methods inherited from class java.lang.Object

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

Field Detail

derivX

protected Derivative extends boofcv.struct.image.ImageBase derivX

derivY

protected Derivative extends boofcv.struct.image.ImageBase derivY

pixelDX

protected float pixelDX

pixelDY

protected float pixelDY

Constructor Detail

DescribeDenseHogAlg

public DescribeDenseHogAlg(int orientationBins,
                           int widthCell,
                           int widthBlock,
                           int stepBlock,
                           boofcv.struct.image.ImageType<Input> imageType)

Configures HOG descriptor computation

Parameters:

orientationBins - Number of bins in a cell's histogram. 9 recommended

widthCell - Number of pixel's wide a cell is. 8 recommended

widthBlock - Number of cells's wide a black is. 3 recommended

stepBlock - Number of cells which are skipped between each block

Method Detail

computeCellWeights

protected void computeCellWeights()

Computes the value of weights inside of a block

setInput

public void setInput(Input input)

Specifies input image. Gradient is computed immediately

Parameters:

input - input image

process

public void process()

Computes the descriptor across the input image

getDescriptorsInRegion

public void getDescriptorsInRegion(int pixelX0,
                                   int pixelY0,
                                   int pixelX1,
                                   int pixelY1,
                                   java.util.List<TupleDesc_F64> output)

Convenience function which returns a list of all the descriptors computed inside the specified region in the image

Parameters:

pixelX0 - Pixel coordinate X-axis lower extent

pixelY0 - Pixel coordinate Y-axis lower extent

pixelX1 - Pixel coordinate X-axis upper extent

pixelY1 - Pixel coordinate Y-axis upper extent

output - List of descriptions

computeDerivative

public abstract void computeDerivative(int pixelIndex)

Computes and stores the gradient at the specified pixel. pixelIndex = y*columns + x

getLocations

public org.ddogleg.struct.FastQueue<georegression.struct.point.Point2D_I32> getLocations()

List of locations for each descriptor.

getDescriptions

public org.ddogleg.struct.FastQueue<TupleDesc_F64> getDescriptions()

List of descriptors

_getDerivX

public Derivative _getDerivX()

_getDerivY

public Derivative _getDerivY()

_getPixelDX

public float _getPixelDX()

_getPixelDY

public float _getPixelDY()

getRegionWidthPixel

public int getRegionWidthPixel()

Returns the number of pixel's wide the square region is that a descriptor was computed from

Returns:

number of pixels wide

setWidthCell

public void setWidthCell(int widthCell)

getWidthCell

public int getWidthCell()

getWidthBlock

public int getWidthBlock()

getStepBlock

public int getStepBlock()

getOrientationBins

public int getOrientationBins()

getCellRows

public int getCellRows()

getCellCols

public int getCellCols()

getCell

public DescribeDenseHogAlg.Cell getCell(int row,
                                        int col)

setStepBlock

public void setStepBlock(int stepBlock)

getImageType

public boofcv.struct.image.ImageType<Input> getImageType()

createDescription

public TupleDesc_F64 createDescription()