boofcv.alg.feature.detect.interest

Class FeatureLaplacePyramid<T extends ImageSingleBand,D extends ImageSingleBand>

java.lang.Object

boofcv.alg.feature.detect.interest.FeatureLaplacePyramid<T,D>

All Implemented Interfaces:

InterestPointScaleSpacePyramid<T>

public class FeatureLaplacePyramid<T extends ImageSingleBand,D extends ImageSingleBand>
extends Object
implements InterestPointScaleSpacePyramid<T>

Feature detector across image pyramids that uses the Laplacian to determine strength in scale-space.

COMMENT ON SCALEPOWER: To normalize feature intensity across scales each feature intensity is multiplied by the scale to the power of 'scalePower'. See [1,2] for how to compute 'scalePower'. Inside of the image pyramid sub-sampling of the image causes the image gradient to be a factor of 'scale' larger than it would be without sub-sampling. In some situations this can negate the need to adjust feature intensity further.

[1] Krystian Mikolajczyk and Cordelia Schmid, "Indexing based on scale invariant interest points" ICCV 2001. Proceedings.
[2] Lindeberg, T., "Feature detection with automatic scale selection." IJCV 30(2) (1998) 79 – 116

Author:

Peter Abeles

See Also:

FactoryInterestPoint

Field Summary

Fields

Modifier and Type	Field and Description
protected AnyImageDerivative<T,D>	computeDerivative
protected List<ScalePoint>	foundPoints
protected List<georegression.struct.point.Point2D_I16>[]	maximums
protected double	scalePower
protected int	spaceIndex

Constructor Summary

Constructors

Constructor and Description
FeatureLaplacePyramid(GeneralFeatureDetector<T,D> detector, ImageFunctionSparse<T> sparseLaplace, AnyImageDerivative<T,D> computeDerivative, double scalePower) Create a feature detector.

Method Summary

Methods

Modifier and Type	Method and Description
void	detect(PyramidFloat<T> ss) Searches for features inside the provided scale space
protected void	findLocalScaleSpaceMax(PyramidFloat<T> ss, int layerID) See if each feature is a maximum in its local scale-space.
List<ScalePoint>	getInterestPoints() Returns all the found interest points

Methods inherited from class java.lang.Object

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

Field Detail

spaceIndex

protected int spaceIndex

maximums

protected List<georegression.struct.point.Point2D_I16>[] maximums

foundPoints

protected List<ScalePoint> foundPoints

computeDerivative

protected AnyImageDerivative<T extends ImageSingleBand,D extends ImageSingleBand> computeDerivative

scalePower

protected double scalePower

Constructor Detail

FeatureLaplacePyramid

public FeatureLaplacePyramid(GeneralFeatureDetector<T,D> detector,
                     ImageFunctionSparse<T> sparseLaplace,
                     AnyImageDerivative<T,D> computeDerivative,
                     double scalePower)

Create a feature detector.

Parameters:

detector - Point feature detector which is used to find candidates in each scale level

sparseLaplace - Used to compute the Laplacian at each candidates

computeDerivative - Used to compute image derivatives

scalePower - Used to normalize features intensity across scale space. For many features this value should be one.

Method Detail

detect

public void detect(PyramidFloat<T> ss)

Searches for features inside the provided scale space

Specified by:

detect in interface InterestPointScaleSpacePyramid<T extends ImageSingleBand>

Parameters:

ss - Scale space of an image

findLocalScaleSpaceMax

protected void findLocalScaleSpaceMax(PyramidFloat<T> ss,
                          int layerID)

See if each feature is a maximum in its local scale-space.

getInterestPoints

public List<ScalePoint> getInterestPoints()

Description copied from interface: InterestPointScaleSpacePyramid

Returns all the found interest points

Specified by:

getInterestPoints in interface InterestPointScaleSpacePyramid<T extends ImageSingleBand>

Returns:

List of found interest points.