boofcv.alg.filter.binary.impl

Class ThresholdLocalSquareBorder

java.lang.Object

boofcv.alg.filter.binary.impl.ThresholdLocalSquareBorder

public class ThresholdLocalSquareBorder
extends java.lang.Object

A locally adaptive thesholding algorithm intended to ensure that shapes which touch the image border are marked as a continuous region. Some locally adaptive algorithms (e.g. mean or gaussian) will break the object's edge when it intersects the image border. This is caused by the lack of an observed edge and the object being textureless. When other some locally adaptive algorithms encounter textureless regions they arbitrary classify pixels. This algorithm will identify that its textureless and mark it as 1.

The algorithm has the following steps:
1) Compute local histogram from square region around the current pixel.
2) Compute the lower and upper intensity values from user configurable percentiles.
3) See if there is sufficient difference between pixel values, if not then it is assumed to be textureless
4a) If textureless mark as 1 4b) If not textureless set a threshold to be the average of the lower and upper values.
If threshold down then 1 = value ≤ threshold. Threshold up, 1 = value > threshold.

Constructor Summary

Constructors

Constructor and Description
ThresholdLocalSquareBorder(boolean thresholdDown, int regionWidth, int histogramLength, int minimumSpread, double lowerFrac, double upperFrac)

Method Summary

Modifier and Type	Method and Description
int	getHistogramLength()
void	process(ImageUInt8 input, ImageUInt8 output)
void	processInner()
void	processSides() Process the border.

Methods inherited from class java.lang.Object

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

Constructor Detail

ThresholdLocalSquareBorder

public ThresholdLocalSquareBorder(boolean thresholdDown,
                                  int regionWidth,
                                  int histogramLength,
                                  int minimumSpread,
                                  double lowerFrac,
                                  double upperFrac)

Method Detail

process

public void process(ImageUInt8 input,
                    ImageUInt8 output)

processInner

public void processInner()

processSides

public void processSides()

Process the border. The histogram is computed by a square flush against the border. Then all the points inside the rectangle "below" the mid point are thresholded using it.

getHistogramLength

public int getHistogramLength()