liop.c File Reference

Local Intensity Order Pattern (LIOP) descriptor - Definition. More...

#include "liop.h"
#include "mathop.h"
#include "imopv.h"
#include <string.h>
#include "qsort-def.h"

Functions
vl_index	get_permutation_index (vl_uindex *permutation, vl_size size)
	Compute permutation index.
VlLiopDesc *	vl_liopdesc_new (vl_int numNeighbours, vl_int numSpatialBins, float radius, vl_size sideLength)
	Create a new LIOP object instance.
VlLiopDesc *	vl_liopdesc_new_basic (vl_size sideLength)
	Create a new object with default parameters.
void	vl_liopdesc_delete (VlLiopDesc *self)
	Delete object instance.
void	vl_liopdesc_process (VlLiopDesc *self, float *desc, float const *patch)
	Compute liop descriptor for a patch.
vl_size	vl_liopdesc_get_dimension (VlLiopDesc const *self)
	Get the dimension of a LIOP descriptor.
vl_size	vl_liopdesc_get_num_neighbours (VlLiopDesc const *self)
	Get the number of neighbours.
float	vl_liopdesc_get_intensity_threshold (VlLiopDesc const *self)
	Get the intensity threshold.
void	vl_liopdesc_set_intensity_threshold (VlLiopDesc *self, float x)
	Set the intensity threshold.
double	vl_liopdesc_get_neighbourhood_radius (VlLiopDesc const *self)
	Get the neighbourhood radius.
vl_size	vl_liopdesc_get_num_spatial_bins (VlLiopDesc const *self)
	Get the number of spatial bins.

---

Detailed Description

Author:

Hana Sarbortova

Andrea Vedaldi

---

Function Documentation

vl_index get_permutation_index ( vl_uindex *  permutation, vl_size  size  )

inline

Parameters:

permutation	array containing all values from 0 to (size - 1) (input/output).
size	size of the permutation array.

Returns:

permutation index.

Compute the position of permutation in the lexycographcial sorting of permutations of the given size.

For example, in the lexicographical ordering, permutations of four elements are listed as [1 2 3 4], [1 2 4 3], [1 3 2 4], [1 3 4 2], [1 4 2 3], [1 4 3 2], [2 1 3 4], ..., [4 3 2 1].

The index can be computed as follows. First pick the first digit perm[1]. This is either 1,2,...,n. For each choice of the first digits, there are (n-1)! other permutations, separated therefore by (n-1)! elements in lexicographical order.

Process then the second digit perm[2]. This can be though as finding the lexycotraphical index of perm[2], ..., perm[n], a permutation of n-1 elements. This can be explicitly obtained by taking out 1 from all elements perm[i] > perm[1].

void vl_liopdesc_delete

(

VlLiopDesc *

self

)

Parameters:

self	object instance.

vl_size vl_liopdesc_get_dimension

(

VlLiopDesc const *

self

)

Returns:

dimension.

float vl_liopdesc_get_intensity_threshold

(

VlLiopDesc const *

self

)

Parameters:

self

object.

Returns:

intensity threshold.

See also:

liop-weighing

double vl_liopdesc_get_neighbourhood_radius

(

VlLiopDesc const *

self

)

Parameters:

self

object.

Returns:

neighbourhood radius.

vl_size vl_liopdesc_get_num_neighbours

(

VlLiopDesc const *

self

)

Parameters:

self

object.

Returns:

number of neighbours.

vl_size vl_liopdesc_get_num_spatial_bins

(

VlLiopDesc const *

self

)

Parameters:

self

object.

Returns:

number of spatial bins.

VlLiopDesc* vl_liopdesc_new	(	vl_int	numNeighbours,
		vl_int	numSpatialBins,
		float	radius,
		vl_size	sideLength
	)

Parameters:

numNeighbours	number of neighbours.
numSpatialBins	number of bins.
radius	radius of the cirucal sample neighbourhoods.
sideLength	width of the input image patch (the patch is square).

Returns:

new object instance.

The value of radius should be at least less than half the sideLength of the patch.

VlLiopDesc* vl_liopdesc_new_basic

(

vl_size

sideLength

)

Parameters:

patchSideLength

size of the patches to be processed.

Returns:

new object.

See also:

vl_liopdesc_new.

void vl_liopdesc_process	(	VlLiopDesc *	self,
		float *	desc,
		float const *	patch
	)

Parameters:

self	object instance
desc	descriptor to be computed (output).
patch	patch to process

Use vl_liopdesc_get_dimension to get the size of the descriptor .

void vl_liopdesc_set_intensity_threshold	(	VlLiopDesc *	self,
		float	x
	)

Parameters:

self	object.
x	intensity threshold.

If non-negative, the threshold as is is used when comparing intensities. If negative, the absolute value of the specified number is multipled by the maximum intensity difference inside a patch to obtain the threshold.

See also:

liop-weighing