boofcv.alg.feature.describe

Class DescribePointSift

java.lang.Object

boofcv.alg.feature.describe.DescribePointSift

public class DescribePointSift
extends Object

Detects SIFT features inside the provided SIFT scale-space. SIFT features work by sampling the image in a grid. For each grid element, a histogram of the gradient's orientation is computed. From this histogram the descriptor is computed. See [1] for the details and below for algorithmic changes.

DESCRIPTOR INTERPOLATION: Instead of using trilinear interpolation a Gaussian weight is used instead. Both methods were tried and Gaussian weight produced slightly better results. Same results had been found by others with regard to SURF descriptors.

[1] Lowe, D. "Distinctive image features from scale-invariant keypoints". International Journal of Computer Vision, 60, 2 (2004), pp.91--110.

Author:

Peter Abeles

Constructor Summary

Constructors

Constructor and Description
DescribePointSift(int gridWidth, int numSamples, int numHistBins, double weightSigma, double sigmaToRadius) Configures detector

Method Summary

Methods

Modifier and Type	Method and Description
int	getDescriptorLength()
void	process(double c_x, double c_y, double scale, double orientation, int imageIndex, double pixelScale, SurfFeature desc) Compute the descriptor with information on which level in the scale-space to use.
void	process(double c_x, double c_y, double scale, double orientation, SurfFeature desc)
void	setScaleSpace(SiftImageScaleSpace ss)

Methods inherited from class java.lang.Object

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

Constructor Detail

DescribePointSift

public DescribePointSift(int gridWidth,
                 int numSamples,
                 int numHistBins,
                 double weightSigma,
                 double sigmaToRadius)

Configures detector

Parameters:

gridWidth - Number of grid elements along a side. Typically 4

numSamples - Number of samples along a grid. Typically 8

numHistBins - Number of bins in the orientation histogram. Typically 8

weightSigma - Adjusts descriptor element's weighting from center. Typically 0.5

sigmaToRadius - Conversation from scale space to pixels. Typically 2.5

Method Detail

setScaleSpace

public void setScaleSpace(SiftImageScaleSpace ss)

process

public void process(double c_x,
           double c_y,
           double scale,
           double orientation,
           SurfFeature desc)

process

public void process(double c_x,
           double c_y,
           double scale,
           double orientation,
           int imageIndex,
           double pixelScale,
           SurfFeature desc)

Compute the descriptor with information on which level in the scale-space to use.

Parameters:

c_x - Location of feature in input image

c_y - Location of feature in input image

scale - Size of the feature in the input image

orientation - Orientation of the feature

imageIndex - Which octave contains the feature

pixelScale - The scale of a pixel in the octave

desc - (Output) storage for the descriptor

getDescriptorLength

public int getDescriptorLength()