Histogram of Oriented Gradients (HOG) - Definition. More...

#include "hog.h"
#include "mathop.h"
#include <string.h>

Functions
VlHog *	vl_hog_new (VlHogVariant variant, vl_size numOrientations, vl_bool transposed)
	Create a new HOG object.
void	vl_hog_delete (VlHog *self)
	Delete a HOG object.
vl_size	vl_hog_get_glyph_size (VlHog const *self)
	Get HOG glyph size.
vl_index const *	vl_hog_get_permutation (VlHog const *self)
	Get HOG left-right flip permutation.
void	vl_hog_set_use_bilinear_orientation_assignments (VlHog *self, vl_bool x)
	Turn bilinear interpolation of assignments on or off.
vl_bool	vl_hog_get_use_bilinear_orientation_assignments (VlHog const *self)
	Tell whether assignments use bilinear interpolation or not.
void	vl_hog_render (VlHog const self, float image, float const *descriptor, vl_size width, vl_size height)
	Render a HOG descriptor to a glyph image.
vl_size	vl_hog_get_dimension (VlHog const *self)
	Get the dimension of the HOG features.
vl_size	vl_hog_get_width (VlHog *self)
	Get the width of the HOG cell array.
vl_size	vl_hog_get_height (VlHog *self)
	Get the height of the HOG cell array.
static void	vl_hog_prepare_buffers (VlHog *self, vl_size width, vl_size height, vl_size cellSize)
	Prepare internal buffers.
void	vl_hog_put_image (VlHog self, float const image, vl_size width, vl_size height, vl_size numChannels, vl_size cellSize)
	Process features starting from an image.
void	vl_hog_put_polar_field (VlHog self, float const modulus, float const *angle, vl_bool directed, vl_size width, vl_size height, vl_size cellSize)
	Process features starting from a field in polar notation.
void	vl_hog_extract (VlHog self, float features)
	Extract HOG features.

Detailed Description

Author:: Andrea Vedaldi

Function Documentation

void vl_hog_delete ( VlHog * self )

Parameters:

self	HOG object to delete.

void vl_hog_extract	(	VlHog *	self,
		float *	features
	)

Parameters:

self	HOG object.
features	HOG features (output).

This method is called after vl_hog_put_image or vl_hog_put_polar_field in order to retrieve the computed HOG features. The buffer features must have the dimensions returned by vl_hog_get_width, vl_hog_get_height, and vl_hog_get_dimension.

vl_size vl_hog_get_dimension ( VlHog const * self )

Parameters:

self	HOG object.

Returns:: imension of a HOG cell descriptors.

vl_size vl_hog_get_glyph_size ( VlHog const * self )

Parameters:

self	HOG object.

Returns:: size (height and width) of a glyph.

vl_size vl_hog_get_height ( VlHog * self )

Parameters:

self	HOG object.

Returns:: number of HOG cells in the vertical direction.

vl_index const* vl_hog_get_permutation ( VlHog const * self )

Parameters:

self	HOG object.

Returns:: left-right permutation.

The function returns a pointer to an array permutation of vl_hog_get_dimension elements. Given a HOG descriptor (for a cell) hog, which is also a vector of vl_hog_get_dimension elements, the descriptor obtained for the same image flipped horizotnally is given by flippedHog[i] = hog[permutation[i]].

vl_bool vl_hog_get_use_bilinear_orientation_assignments ( VlHog const * self )

Parameters:

self	HOG object.

Returns:: true if orientations are be assigned with bilinear interpolation.

vl_size vl_hog_get_width ( VlHog * self )

Parameters:

self	HOG object.

Returns:: number of HOG cells in the horizontal direction.

VlHog* vl_hog_new	(	VlHogVariant	variant,
		vl_size	numOrientations,
		vl_bool	transposed
	)

Parameters:

variant	HOG descriptor variant.
numOrientations	number of distinguished orientations.
transposed	wether images are transposed (column major).

Returns:: the new HOG object.

The function creates a new HOG object to extract descriptors of the prescribed variant. The angular resolution is set by numOrientations, which specifies the number of undirected orientations. The object can work with column major images by setting transposed to true.

static void vl_hog_prepare_buffers	(	VlHog *	self,
		vl_size	width,
		vl_size	height,
		vl_size	cellSize
	)

static

Parameters:

self	HOG object.
width	image width.
height	image height.
cellSize	size of a HOG cell.

void vl_hog_put_image	(	VlHog *	self,
		float const *	image,
		vl_size	width,
		vl_size	height,
		vl_size	numChannels,
		vl_size	cellSize
	)

Parameters:

self	HOG object.
image	image to process.
width	image width.
height	image height.
numChannels	number of image channles.
cellSize	size of a HOG cell.

The buffer hog must be a three-dimensional array. The first two dimensions are (width + cellSize/2)/cellSize and (height + cellSize/2)/cellSize, where divisions are integer. This is approximately width/cellSize and height/cellSize, adjusted so that the last cell is at least half contained in the image.

The image width and height must be not smaller than three pixels and not smaller than cellSize.

void vl_hog_put_polar_field	(	VlHog *	self,
		float const *	modulus,
		float const *	angle,
		vl_bool	directed,
		vl_size	width,
		vl_size	height,
		vl_size	cellSize
	)

Parameters:

self	HOG object.
modulus	image gradient modulus.
angle	image gradient angle.
directed	wrap the gradient angles at 2pi (directed) or pi (undirected).
width	image width.
height	image height.
cellSize	size of a HOG cell.

The function behaves like vl_hog_put_image, but foregoes the internal computation of the gradient field, allowing the user to specify their own. Angles are measure clockwise, the y axis pointing downwards, starting from the x axis (pointing to the right).

void vl_hog_render	(	VlHog const *	self,
		float *	image,
		float const *	descriptor,
		vl_size	width,
		vl_size	height
	)

Parameters:

self	HOG object.
image	glyph image (output).
descriptor	HOG descriptor.
width	HOG descriptor width.
height	HOG descriptor height.

The function renders the HOG descriptor or filter descriptor as an image (for visualization) and stores the result in the buffer image. This buffer must be an array of dimensions width*glyphSize by height*glyphSize elements, where glyphSize is obtained from vl_hog_get_glyph_size and is the size in pixels of the image element used to represent the descriptor of one HOG cell.

void vl_hog_set_use_bilinear_orientation_assignments	(	VlHog *	self,
		vl_bool	x
	)

Parameters:

self	HOG object.
x	`true` if orientations should be assigned with bilinear interpolation.

VLFeat.org

Functions

Detailed Description

Function Documentation