Image Analysis with flow and gray scale

Hello. I am trying to calculate the percentage of the white foam for this image. I am struggling because I can either do a 8 bit binary black and white but it takes out some of the foam that is less dense and not completely white. The color of the foam varies. The second way I approached it was using a RBG analysis. I basically just analyzed the closest ranges for gray to white and black as the background.

Do any of you know how to separate the foam from the background more accurately?

Thanks

   I = imread('test5.png') ;
  figure; 
  imshow(I)
  BW = im2bw(I,0.4);
  figure; imshow(BW)
  percentageBlack=(1-nnz(BW)/numel(BW))*100

%%%%another method

clear all
clc

filename = 'test5.png';
RGB = imread(filename);
grayscale = rgb2gray(RGB);

numTotalPixel = size(grayscale,1) * size(grayscale, 2);
numBlackPixel = sum(grayscale(:)<= 80);
numWhitePixel = sum(grayscale(:)>140);
numGrayPixel  = sum(grayscale(:)>80)-numWhitePixel;

percentBlackPixel = numBlackPixel / numTotalPixel * 100
percentWhitePixel = numWhitePixel / numTotalPixel * 100
percentGrayPixel = numGrayPixel / numTotalPixel *100

Answer 1

[Image Analyst]

I have done foam/suds/lather analysis. If a gray level (like 100) might be foam in one location, but background in another location, then you're going to have to come up with some descriptor of foam that depends on more than simply the gray scale. I don't know what that might be. Maybe you need to do some spatial processing first, like using adapthisteq() or something.

[Christine Nee]

How did you approach it in the past? Do you have any examples or advice? Thanks

[Image Analyst]

In my case we just thresholded at different gray levels to find thick, thin, or no foam/bubbles. But our foam could be found simply on the basis of gray level.

[Christine Nee]

Then you would go back and do a pixel analysis on it?

[Image Analyst]

I filled in the clear bubbles using imclose(). Then I thresholded, smoothed the boundaries a bit I think, and finally got area fractions from the histogram.