boofcv.alg.feature.detect.line

Class HoughTransformLinePolar

java.lang.Object

boofcv.alg.feature.detect.line.HoughTransformLinePolar

public class HoughTransformLinePolar
extends java.lang.Object

Hough transform which uses a polar line representation, distance from origin and angle (0 to 180 degrees). Standard implementation of a hough transform. 1) Gradient intensity image is used to find edge pixels. 2) All possible lines passing through that point are found. 3) Line parameters are summed up in the line image, in which each pixel represents a coordinate in parameter space. 3) Local maximums are found.

By the nature of this algorithms, lines are forced to be discretized into parameter space. The range can vary from +- the maximum range inside the image and the angle from 0 to PI radians. How finely discretized an image is effects line detection accuracy. If too fine lines might not be detected or it will be too noisy.

In the line image, the transform from line parameter coordinate to pixel coordinate is as follow:
x = r*cos(theta) + centerX
y = r*sin(theta) + centerY

USAGE NOTE: Duplicate/very similar lines are possible due to angles being cyclical. What happens is that if a line's orientation lies along a boundary point its angles will be split up between top and bottom of the transform. When lines are extracted using non-maximum it will detects peaks at the top and bottom.

Constructor Summary

Constructors

Constructor and Description
HoughTransformLinePolar(NonMaxSuppression extractor, int numBinsRange, int numBinsAngle) Specifies parameters of transform.

Method Summary

Modifier and Type	Method and Description
FastQueue<LineParametric2D_F32>	extractLines() Searches for local maximas and converts into lines.
float[]	getFoundIntensity() Returns the intensity/edge count for each returned line.
int	getNumBinsAngle()
int	getNumBinsRange()
GrayF32	getTransform() // * Returns the Hough transform image.
void	lineToCoordinate(LineParametric2D_F32 line, Point2D_F64 coordinate) Compute the parameterized coordinate for the line
void	parameterize(int x, int y) Converts the pixel coordinate into a line in parameter space
void	transform(GrayU8 binary) Computes the Hough transform of the image.

Methods inherited from class java.lang.Object

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

Constructor Detail

HoughTransformLinePolar

public HoughTransformLinePolar(NonMaxSuppression extractor,
                               int numBinsRange,
                               int numBinsAngle)

Specifies parameters of transform. The minimum number of points specified in the extractor is an important tuning parameter.

Parameters:

extractor - Extracts local maxima from transform space.

numBinsRange - How many bins are be used for line range.

numBinsAngle - How many bins are used for angle.

Method Detail

getNumBinsRange

public int getNumBinsRange()

getNumBinsAngle

public int getNumBinsAngle()

transform

public void transform(GrayU8 binary)

Computes the Hough transform of the image.

Parameters:

binary - Binary image that indicates which pixels lie on edges.

extractLines

public FastQueue<LineParametric2D_F32> extractLines()

Searches for local maximas and converts into lines.

Returns:

Found lines in the image.

lineToCoordinate

public void lineToCoordinate(LineParametric2D_F32 line,
                             Point2D_F64 coordinate)

Compute the parameterized coordinate for the line

parameterize

public void parameterize(int x,
                         int y)

Converts the pixel coordinate into a line in parameter space

getTransform

public GrayF32 getTransform()

// * Returns the Hough transform image.

Returns:

Transform image.

getFoundIntensity

public float[] getFoundIntensity()

Returns the intensity/edge count for each returned line. Useful when doing post processing pruning.

Returns:

Array containing line intensities.