Package cz.vutbr.fit.layout.cormier.impl

Class EdgeDetector

java.lang.Object

cz.vutbr.fit.layout.cormier.impl.EdgeDetector

public class EdgeDetector
extends Object

Detector for locally significant edges for the CormierSegmentation.

Author:

František Maštera

See Also:

Michael Cormier (2018). Computer Vision on Web Pages: A Study of Man-Made Images. UWSpace.

Constructor Summary

Constructors

Constructor	Description
EdgeDetector(int halfWindowWidth, float standardDeviation, float priorProbability, int pyramidLevels)
EdgeDetector(int halfWindowWidth, float standardDeviation, float priorProbability, int pyramidLevels, boolean ndPregenEnabled, boolean parallelEnabled)

Method Summary

Modifier and Type	Method	Description
org.javatuples.Pair<double[][],double[][]>	getEdgeProbabilities(byte[] imageData)
org.javatuples.Pair<double[][],double[][]>	getEdgeProbabilities(org.opencv.core.Mat image)
int	getHalfWindowWidth()
float	getPriorProbability()
int	getPyramidLevels()
float	getStandardDeviation()
org.javatuples.Pair<org.opencv.core.Mat,org.opencv.core.Mat>	multiscaleSobel(org.opencv.core.Mat greyscaleImage)
void	setHalfWindowWidth(int halfWindowWidth)
void	setPriorProbability(float priorProbability)
void	setPyramidLevels(int pyramidLevels)
void	setStandardDeviation(float standardDeviation)

Methods inherited from class java.lang.Object

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

Constructor Detail

EdgeDetector

public EdgeDetector(int halfWindowWidth,
                    float standardDeviation,
                    float priorProbability,
                    int pyramidLevels)

Parameters:

halfWindowWidth - Range of the neighborhood (around the pixel) taken into account when applying kernel density estimation using a Gaussian kernel.

standardDeviation - Standard deviation used when applying kernel density estimation using a Gaussian kernel.

priorProbability - Prior probability that a given pixel is an edge.

pyramidLevels - Number of levels for the Gaussian pyramid used for the multiscale edge detection. Set to 1 to effectively turn off the multiscale edge detection

EdgeDetector

public EdgeDetector(int halfWindowWidth,
                    float standardDeviation,
                    float priorProbability,
                    int pyramidLevels,
                    boolean ndPregenEnabled,
                    boolean parallelEnabled)

Parameters:

ndPregenEnabled - If the edgeProbabilities(Mat, Mat) optimization by pre-generating and storing the NormalDistribution instances will be enabled. Disabling this will reduce the memory requirements significantly, but also slows down the process.

parallelEnabled - If the edgeProbabilities(Mat, Mat) parallelization will be enabled. Disabling this will cause significant slow down on multithreaded systems.

See Also:

EdgeDetector(int, float, float, int)

Method Detail

getHalfWindowWidth

public int getHalfWindowWidth()

Returns:

Range of the neighborhood (around the pixel) taken into account when applying kernel density estimation using a Gaussian kernel.

setHalfWindowWidth

public void setHalfWindowWidth(int halfWindowWidth)

Parameters:

halfWindowWidth - Range of the neighborhood (around the pixel) taken into account when applying kernel density estimation using a Gaussian kernel.

getStandardDeviation

public float getStandardDeviation()

Returns:

Standard deviation used when applying kernel density estimation using a Gaussian kernel.

setStandardDeviation

public void setStandardDeviation(float standardDeviation)

Parameters:

standardDeviation - Standard deviation used when applying kernerl density estimation using a Gaussian kernel.

getPriorProbability

public float getPriorProbability()

Returns:

Prior probability that a given pixel is an edge.

setPriorProbability

public void setPriorProbability(float priorProbability)

Parameters:

priorProbability - Prior probability that a given pixel is an edge.

getPyramidLevels

public int getPyramidLevels()

Returns:

Number of levels for the Gaussian pyramid used for the multiscale edge detection. Set to 1 to effectively turn off the multiscale edge detection

setPyramidLevels

public void setPyramidLevels(int pyramidLevels)

Parameters:

pyramidLevels - Number of levels for the Gaussian pyramid used for the multiscale edge detection. Set to 1 to effectively turn off the multiscale edge detection

getEdgeProbabilities

public org.javatuples.Pair<double[][],double[][]> getEdgeProbabilities(byte[] imageData)

Returns:

Probabilities of a horizontal/vertical edge on each pixel in the given image.

getEdgeProbabilities

public org.javatuples.Pair<double[][],double[][]> getEdgeProbabilities(org.opencv.core.Mat image)

Returns:

Probabilities of a horizontal/vertical edge on each pixel in the given image.

multiscaleSobel

public org.javatuples.Pair<org.opencv.core.Mat,org.opencv.core.Mat> multiscaleSobel(org.opencv.core.Mat greyscaleImage)

Returns:

Horzontal and vertical edges for the given image, calculated using multiscale Imgproc.Sobel(Mat, Mat, int, int, int) edge detection with a Gaussian pyramid with getPyramidLevels() levels. Calling sobel(Mat) returns the same result as calling this method when getPyramidLevels() is equal to 1. The dimensions of both horizontal and vertical edges are the same as the original image.