boofcv.abst.feature.detect.extract

Interface NonMaxSuppression

All Known Implementing Classes:

WrapperNonMaxCandidate, WrapperNonMaximumBlock, WrapperNonMaximumNaive

public interface NonMaxSuppression

Detects local minimums and/or maximums in an intensity image inside square regions. This is known as non-maximum suppression. The detector can be configured to ignore pixels along the image border by a user specified distance. Some implementations require candidate locations for the features. This allows for a sparse algorithm to be used, resulting in a significant speed boost. Pixel values with a value of -Float.MAX_VALUE or Float.MAX_VALUE will not be considered for local minimum/maximum, respectively. This is a good way to ignore previously detected features.

Not all implementations will search for both minimums or maximums. Be sure you are using the correct one. If you don't intend on detecting a minimum or maximum pass in null for the candidate list and the output found list.

An extractor which uses candidate features must always be provided them. However, an algorithm which does not use candidate features will simply ignore that input and operate as usual. Can check capabilities at runtime using the canDetectMinimums() and canDetectMaximums() functions.

The processing border is defined as the image minus the ignore border. Some algorithms cannot detect features which are within the search radius of this border. If that is the case it would be possible to have a feature at the image border. To determine if this is the case call canDetectBorder().

Method Summary

Methods

Modifier and Type	Method and Description
boolean	canDetectBorder() Can it detect features which are inside the image border.
boolean	canDetectMaximums() True if it can detect local maximums.
boolean	canDetectMinimums() True if it can detect local minimums.
int	getIgnoreBorder() Returns the size of the image border which is not processed.
int	getSearchRadius() Describes how large the region is that is being searched.
float	getThresholdMaximum() Minimum value for detected maximums
float	getThresholdMinimum() Maximum value for detected minimums
boolean	getUsesCandidates() Returns true if the algorithm requires a candidate list of corners.
void	process(boofcv.struct.image.ImageFloat32 intensity, boofcv.struct.QueueCorner candidateMin, boofcv.struct.QueueCorner candidateMax, boofcv.struct.QueueCorner foundMin, boofcv.struct.QueueCorner foundMax) Process a feature intensity image to extract the point features.
void	setIgnoreBorder(int border) Pixels which are within 'border' of the image border will not be considered.
void	setSearchRadius(int radius) Species the search radius for the feature
void	setThresholdMaximum(float threshold) Change the feature selection threshold for finding local maximums.
void	setThresholdMinimum(float threshold) Change the feature selection threshold for finding local minimums.

Method Detail

process

void process(boofcv.struct.image.ImageFloat32 intensity,
           boofcv.struct.QueueCorner candidateMin,
           boofcv.struct.QueueCorner candidateMax,
           boofcv.struct.QueueCorner foundMin,
           boofcv.struct.QueueCorner foundMax)

Process a feature intensity image to extract the point features. If a pixel has an intensity value == -Float.MAX_VALUE or Float.MAX_VALUE it will not be considered for a local min or max, respectively. If an algorithm only detect local minimums or maximums and null can be passed in for unused lists. This is the recommended procedure since it will force an exception to be thrown if a mistake was made.

Parameters:

intensity - (Input) Feature intensity image. Not modified.

candidateMin - (Input) (Optional) List of candidate local minimum features. Can be null if not used.

candidateMax - (Input) (Optional) List of candidate local maximum features Can be null if not used.

foundMin - (Output) Storage for found minimums. Can be null if not used.

foundMax - (Output) Storage for found maximums. Can be null if not used.

getUsesCandidates

boolean getUsesCandidates()

Returns true if the algorithm requires a candidate list of corners.

Returns:

true if candidates are required.

getThresholdMinimum

float getThresholdMinimum()

Maximum value for detected minimums

Returns:

threshold for feature selection

getThresholdMaximum

float getThresholdMaximum()

Minimum value for detected maximums

Returns:

threshold for feature selection

setThresholdMinimum

void setThresholdMinimum(float threshold)

Change the feature selection threshold for finding local minimums.

Parameters:

threshold - The new selection threshold.

setThresholdMaximum

void setThresholdMaximum(float threshold)

Change the feature selection threshold for finding local maximums.

Parameters:

threshold - The new selection threshold.

setIgnoreBorder

void setIgnoreBorder(int border)

Pixels which are within 'border' of the image border will not be considered.

Parameters:

border - Border size in pixels.

getIgnoreBorder

int getIgnoreBorder()

Returns the size of the image border which is not processed.

Returns:

border size

canDetectBorder

boolean canDetectBorder()

Can it detect features which are inside the image border. For example if a feature has a radius of 5, but there is a local max at 2, should that be returned? Can the output handle feature descriptors which are partially inside the image?

Returns:

If it can detect features inside the image border.

setSearchRadius

void setSearchRadius(int radius)

Species the search radius for the feature

Parameters:

radius - Radius in pixels

getSearchRadius

int getSearchRadius()

Describes how large the region is that is being searched. The radius is the number of pixels away from the center.

Returns:

Search radius

canDetectMaximums

boolean canDetectMaximums()

True if it can detect local maximums.

canDetectMinimums

boolean canDetectMinimums()

True if it can detect local minimums.