![JAR search and dependency download from the Maven repository](/logo.png)
net.sourceforge.plantuml.quantization.MedianCutQuantizer Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of plantuml-mit Show documentation
Show all versions of plantuml-mit Show documentation
PlantUML is a component that allows to quickly write diagrams from text.
// THIS FILE HAS BEEN GENERATED BY A PREPROCESSOR.
package net.sourceforge.plantuml.quantization;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collection;
import java.util.Collections;
import java.util.Comparator;
import java.util.HashSet;
import java.util.List;
import java.util.Set;
import java.util.TreeSet;
/**
* Implements median cut quantization.
*
*
* The algorithm works as follows:
*
*
* - Begin with one cluster containing all the original colors.
* - Find the cluster containing the greatest spread along a single color
* component (red, green or blue).
* - Find the median of that color component among colors in the cluster.
* - Split the cluster into two halves, using that median as a threshold.
* - Repeat this process until the desired number of clusters is reached.
*
*/
public final class MedianCutQuantizer implements ColorQuantizer {
public static final MedianCutQuantizer INSTANCE = new MedianCutQuantizer();
private MedianCutQuantizer() {
}
@Override
public Set quantize(Multiset originalColors, int maxColorCount) {
TreeSet clusters = new TreeSet<>(new ClusterSpreadComparator());
clusters.add(new Cluster(originalColors));
while (clusters.size() < maxColorCount) {
Cluster clusterWithLargestSpread = clusters.pollFirst();
clusters.addAll(clusterWithLargestSpread.split());
}
Set clusterCentroids = new HashSet<>();
for (Cluster cluster : clusters) {
clusterCentroids.add(QColor.getCentroid(cluster.colors));
}
return clusterCentroids;
}
private static final class Cluster {
final Multiset colors;
double largestSpread;
int componentWithLargestSpread;
Cluster(Multiset colors) {
this.colors = colors;
this.largestSpread = -1;
for (int component = 0; component < 3; ++component) {
double componentSpread = getComponentSpread(component);
if (componentSpread > largestSpread) {
largestSpread = componentSpread;
componentWithLargestSpread = component;
}
}
}
double getComponentSpread(int component) {
double min = Double.POSITIVE_INFINITY;
double max = Double.NEGATIVE_INFINITY;
for (QColor color : colors) {
min = Math.min(min, color.getComponent(component));
max = Math.max(max, color.getComponent(component));
}
return max - min;
}
Collection split() {
List orderedColors = new ArrayList<>(colors);
Collections.sort(orderedColors, new ColorComponentComparator(componentWithLargestSpread));
int medianIndex = orderedColors.size() / 2;
return Arrays.asList(new Cluster(new HashMultiset<>(orderedColors.subList(0, medianIndex))),
new Cluster(new HashMultiset<>(orderedColors.subList(medianIndex, orderedColors.size()))));
}
}
/**
* Orders clusters according to their maximum spread, in descending order.
*/
static final class ClusterSpreadComparator implements Comparator {
@Override
public int compare(Cluster a, Cluster b) {
double spreadDifference = b.largestSpread - a.largestSpread;
if (spreadDifference == 0) {
return ArbitraryComparator.INSTANCE.compare(a, b);
}
return (int) Math.signum(spreadDifference);
}
}
/**
* Orders colors according to the value of one particular component, in
* ascending order.
*/
static final class ColorComponentComparator implements Comparator {
final int component;
ColorComponentComparator(int component) {
this.component = component;
}
@Override
public int compare(QColor a, QColor b) {
double componentDifference = a.getComponent(component) - b.getComponent(component);
if (componentDifference == 0) {
return ArbitraryComparator.INSTANCE.compare(a, b);
}
return (int) Math.signum(componentDifference);
}
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy