Histogram.java

package uk.co.mmscomputing.image.operators;

public class Histogram{

  int[]  histogram;
  int    N=0;

  Histogram(int slots){
    histogram=new int[slots];
  }

  void write(int colour){
    histogram[colour]++;
    N++;
  }

  /*
	[1] Shunji Mori; Optical Character Recognition (1999); John Wiley & Sons
    	4.2 Thresholding selection based on global discriminant analysis; pp.112

  [2] Ioannis Pavlidis, Vassilios Morellas, Pete Roeber; Programming Cameras and Pan-Tilts with DirectX and Java
      Morgan-Kaufmann Publishers; ISBN 1-55860-756-0
      pp.69

  Original Paper:

  [3] Otsu,N.A Threshold Selection Method from Gray-Level Histograms.
      IEEE Transactions on Systems,Man, and Cybernetics,1979,vol.9,pp.62-66
  */

  int getThreshold(){
    double[] p=new double[histogram.length];

    for(int i=0;i<histogram.length;i++){              // probabilities
      p[i]=((double)histogram[i])/((double)N);
    }
    double mu=0.0;                                    // total mean value
    for(int i=0;i<p.length;i++){mu+=(i+1)*p[i];}

    int    threshold=0;
    double w1=0.0;
    double mu1=0.0;
    double sigmaBSqr;
    double sigmaBSqrMax=0.0;

    for(int i=0;i<p.length;i++){
      w1  +=      p[i];
      mu1 +=(i+1)*p[i];
      if((0<w1)&&(w1<1)){
        sigmaBSqr=mu*w1-mu1;
        sigmaBSqr=(sigmaBSqr*sigmaBSqr)/(w1*(1.0-w1));
        if(sigmaBSqr>sigmaBSqrMax){
          sigmaBSqrMax=sigmaBSqr;
          threshold=i;
        }
      }
    }      
    return threshold;
  }
}