Can anyone help me to find "Maximal Correlation Coefficient"?(Square root of second largest eigenvalue) This is the 14th feature of the Harlick Texture Features. I just need to calculate it out of those features. Can i get a piece of code for it?

a=rgb2gray(imread('IM_0524-Frame1.jpg'));

   a=im2double(a);
   counter=1;

[row,col]=size(a);

for i =2:1:2

      for j=2:1:2
          temp= a(i-1:i+1, j-1:j+1);
          offset1 = [0 0];
          offset2 = [1 1];
          offset3 = [0 1];
          offset4 = [-1 1];
          %glcm1=graycomatrix(temp,'numlevels',256,'offset',[0 1]);
          glcm1 = graycomatrix(temp,'offset',offset1);
         % Need to calculate Maximal correlation coefficient for glcm1.

      end;
  end;

[Walter Roberson]

eigs() ?

[Hanbrot]

Do you mean i need to get the second largest value out of these 3 complex values(according to my case because i am working on 3x3 block) and take its square root? that will give me the required output?

[Bryan Redd]

Did you ever figure this out? I am also looking for the code to calculate the fourteenth feature. I would like to check my work.

[Walter Roberson]

Is it "second largest" according to absolute value? What if there is a tie?

temp = eigs(A, [], 2, 'largestabs');
result = sqrt(temp);

Answer 1

[Pawel Badura]

If you use dr Pierce's PORTS package (https://www.mathworks.com/matlabcentral/fileexchange/55587-ports-3d-image-texture-metric-calculation-package?s_tid=prof_contriblnk), just update the compute_GTSDM_metrics.m script by adding the following code below line 258:

% (14) Maximal Correlation Coefficient
%%% I don't think we use it, so I'll only code it up if needed.
%% ADDED BY PB: 2022-09-30 %%%
Q = zeros(size(p));
p_x0 = max(p_x, 1e-10);     % just to make sure there's no division-by-0 later
p_y0 = max(p_y, 1e-10);     % just to make sure there's no division-by-0 later
for i=1:N_g
    for j=1:N_g
        Q(i,j) = sum((p(i,:) .* p(j,:)) ./ (p_x0(i) * p_y0'));
    end
end
MCC = eigs(Q, 2);
metrics_vect(14) = sqrt(MCC(2));

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%