boofcv.alg.feature.dense

Class DescribeDenseSiftAlg<D extends boofcv.struct.image.ImageGray>

java.lang.Object

boofcv.alg.feature.describe.DescribeSiftCommon

boofcv.alg.feature.dense.DescribeDenseSiftAlg<D>

public class DescribeDenseSiftAlg<D extends boofcv.struct.image.ImageGray>
extends DescribeSiftCommon

Computes SIFT features in a regular grid across an entire image at a single scale and orientation. This is more computationally efficient than the more generic DescribePointSift algorithm because it makes strong assumptions. If given the same center point, an orientation of 0, and sigmaToPixels is 1, they should produce the same descriptor.

Sampling is done in regular increments in a grid pattern. The example sampling points are computed such that entire area sampled starts and ends at the most extreme possible pixels. This most likely will require that the sampling period be adjusted. Multiple descriptors can overlap the same area, so pixel orientation and magnitude is just computed once and saved.

Field Summary

Fields inherited from class boofcv.alg.feature.describe.DescribeSiftCommon

gaussianWeight, histogramBinWidth, maxDescriptorElementValue, numHistogramBins, widthGrid, widthSubregion

Constructor Summary

Constructors

Constructor and Description
DescribeDenseSiftAlg(int widthSubregion, int widthGrid, int numHistogramBins, double weightingSigmaFraction, double maxDescriptorElementValue, double periodColumns, double periodRows, java.lang.Class<D> derivType) Specifies SIFT descriptor structure and sampling frequency.

Method Summary

Modifier and Type	Method and Description
void	computeDescriptor(int cx, int cy, TupleDesc_F64 desc) Computes the descriptor centered at the specified coordinate
java.lang.Class<D>	getDerivType()
org.ddogleg.struct.FastQueue<TupleDesc_F64>	getDescriptors()
org.ddogleg.struct.FastQueue<georegression.struct.point.Point2D_I32>	getLocations() Returns where in the image it sampled the features
double	getPeriodColumns()
double	getPeriodRows()
void	process() Computes SIFT descriptors across the entire image
void	setImageGradient(D derivX, D derivY) Sets the gradient and precomputes pixel orientation and magnitude
void	setPeriodColumns(double periodColumns)
void	setPeriodRows(double periodRows)

Methods inherited from class boofcv.alg.feature.describe.DescribeSiftCommon

createGaussianWeightKernel, getCanonicalRadius, getDescriptorLength, normalizeDescriptor, trilinearInterpolation

Methods inherited from class java.lang.Object

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

Constructor Detail

DescribeDenseSiftAlg

public DescribeDenseSiftAlg(int widthSubregion,
                            int widthGrid,
                            int numHistogramBins,
                            double weightingSigmaFraction,
                            double maxDescriptorElementValue,
                            double periodColumns,
                            double periodRows,
                            java.lang.Class<D> derivType)

Specifies SIFT descriptor structure and sampling frequency.

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

periodColumns - Number of pixels between samples along x-axis

periodRows - Number of pixels between samples along y-axis

derivType - Type of input derivative image

Method Detail

setImageGradient

public void setImageGradient(D derivX,
                             D derivY)

Sets the gradient and precomputes pixel orientation and magnitude

Parameters:

derivX - image derivative x-axis

derivY - image derivative y-axis

process

public void process()

Computes SIFT descriptors across the entire image

computeDescriptor

public void computeDescriptor(int cx,
                              int cy,
                              TupleDesc_F64 desc)

Computes the descriptor centered at the specified coordinate

Parameters:

cx - center of region x-axis

cy - center of region y-axis

desc - The descriptor

getPeriodRows

public double getPeriodRows()

setPeriodRows

public void setPeriodRows(double periodRows)

getPeriodColumns

public double getPeriodColumns()

setPeriodColumns

public void setPeriodColumns(double periodColumns)

getDescriptors

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

getLocations

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

Returns where in the image it sampled the features

getDerivType

public java.lang.Class<D> getDerivType()