Image segmentation using thresholding

I have raw image of rocks like the one on the right, I want a code to get the result on the left

[Cris LaPierre]

What do the different colors correspond to?

Have you tried using the Color Threshholder app? Here's an example that you might find useful

• Segment Image and Create Mask Using Color Thresholder

[Aisha Al Ghurabi]

Rlue : Pores

Red: Grains

Green: in between

[DGM]

Hmm. There may be some better way of doing the classification, but I don't know what it would be. The gray levels that I estimated from the example image seems fairly plausible, though it might be worth trying to adjust them. I am also not really sure what subsequent steps need to be performed on these labeled regions; consequently, I'm not really sure how exact they need to be.

Answer 1

[DGM]

Since we're playing guessing games, here's my guess.

% you might have an actual image, but all we have is a screenshot
inpict = imread('image.bmp'); 
inpict = imcrop(inpict,[3.51 23.51 703.98 471.98]); % crop off the border

% split the two halves
A = im2gray(inpict(:,352+1:end,:)); % input
B = inpict(:,1:352,:); % output

% look at the hue of the output sample
[H,~,~] = rgb2hsv(B);
H = mod(H+0.2,1); % rotate to keep the red peak from wrapping across zero
imhist(H)

% split the histogram between the peaks
mkr = H < 0.25;
mkb = H > 0.80;
mkg = H > 0.45 & H < 0.6;

% sample the source in those regions
Ar = A(mkr);
Ag = A(mkg);
Ab = A(mkb);
subplot(3,1,1), imhist(Ar)
subplot(3,1,2), imhist(Ag)
subplot(3,1,3), imhist(Ab)

% the bins actually overlap quite a bit
% we don't know if that's due to scaling/registration errors
% or whether the classification is done differently
levels = [90 180]; % pick some bin edges
CT = [0 0 1; 0 1 0; 1 0 0]; % a color table
indpict = imquantize(A,levels); % an indexed image

figure
imshow(indpict,CT) % show it