boofcv.alg.feature.describe

Class DescribeSiftCommon

java.lang.Object

boofcv.alg.feature.describe.DescribeSiftCommon

Direct Known Subclasses:

DescribeDenseSiftAlg, DescribePointSift

public class DescribeSiftCommon
extends java.lang.Object

Base class for SIFT descriptors. Provides common functionality between sparse and dense computations.

Field Summary

Fields

Modifier and Type	Field and Description
protected float[]	gaussianWeight
protected double	histogramBinWidth
protected double	maxDescriptorElementValue
protected int	numHistogramBins
protected int	widthGrid
protected int	widthSubregion

Constructor Summary

Constructors

Constructor and Description
DescribeSiftCommon(int widthSubregion, int widthGrid, int numHistogramBins, double weightingSigmaFraction, double maxDescriptorElementValue) Configures the descriptor.

Method Summary

Modifier and Type	Method and Description
protected static float[]	createGaussianWeightKernel(double sigma, int radius) Creates a gaussian weighting kernel with an even number of elements along its width
int	getCanonicalRadius() Radius of descriptor in pixels.
int	getDescriptorLength() Number of elements in the descriptor.
static void	normalizeDescriptor(TupleDesc_F64 descriptor, double maxDescriptorElementValue) Adjusts the descriptor.
protected void	trilinearInterpolation(float weight, float sampleX, float sampleY, double angle, TupleDesc_F64 descriptor) Applies trilinear interpolation across the descriptor

Methods inherited from class java.lang.Object

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

Field Detail

widthSubregion

protected int widthSubregion

widthGrid

protected int widthGrid

numHistogramBins

protected int numHistogramBins

histogramBinWidth

protected double histogramBinWidth

maxDescriptorElementValue

protected double maxDescriptorElementValue

gaussianWeight

protected float[] gaussianWeight

Constructor Detail

DescribeSiftCommon

public DescribeSiftCommon(int widthSubregion,
                          int widthGrid,
                          int numHistogramBins,
                          double weightingSigmaFraction,
                          double maxDescriptorElementValue)

Configures the descriptor.

Parameters:

widthSubregion - Width of sub-region in samples. Try 4

widthGrid - Width of grid in subregions. Try 4.

numHistogramBins - Number of bins in histogram. Try 8

weightingSigmaFraction - Sigma for Gaussian weighting function is set to this value * region width. Try 0.5

maxDescriptorElementValue - Helps with non-affine changes in lighting. See paper. Try 0.2

Method Detail

normalizeDescriptor

public static void normalizeDescriptor(TupleDesc_F64 descriptor,
                                       double maxDescriptorElementValue)

Adjusts the descriptor. This adds lighting invariance and reduces the affects of none-affine changes in lighting. 1) Apply L2 normalization 2) Clip using max descriptor value 3) Apply L2 normalization again

createGaussianWeightKernel

protected static float[] createGaussianWeightKernel(double sigma,
                                                    int radius)

Creates a gaussian weighting kernel with an even number of elements along its width

trilinearInterpolation

protected void trilinearInterpolation(float weight,
                                      float sampleX,
                                      float sampleY,
                                      double angle,
                                      TupleDesc_F64 descriptor)

Applies trilinear interpolation across the descriptor

getDescriptorLength

public int getDescriptorLength()

Number of elements in the descriptor.

getCanonicalRadius

public int getCanonicalRadius()

Radius of descriptor in pixels. Width is radius*2