What is the best way to find an accurate threshold value for im2bw?

I have been attempting to find the best way to threshold an image for the im2bw function however any method I have attempted has resulted in failure. Most recently this has involved plotting a histogram of each colour and depending on where the noise is, using an if loop to remove that noise. The code below demonstrates how I have done this. Does anyone know of a better method so I can accurately find the threshold limit?

if sum(lower_values_red) < sum(higher_values_red) %this loop will decide what part of the histogram to remove

      red_lower_limit = double(mean(r(:))) - double(std(double(r(:))));
      red_upper_limit = red_realogram.BinLimits(2);

else
    sum(lower_values_red) > sum(higher_values_red);

      red_lower_limit = red_realogram.BinLimits(1);
      red_upper_limit = double(mean(r(:))) + double(std(double(r(:))));
  end 

red_thresh = red_lower_limit/red_upper_limit;

Answer 1

[Thorsten]

Steve's blog may help: http://blogs.mathworks.com/steve/

[Nathan Costin]

This has been helpful, I'm currently trying to make it work, specifically used this link:

http://blogs.mathworks.com/steve/2016/06/14/image-binarization-otsus-method/