Subscripted assignment dimension mismatch error

am working on an image processing project and i am getting an error on line 35. the error is Subscripted assignment dimension mismatch error. i have tried to fix it in vain. could someone help me please

1. %D0=25x25 image(double)
2. %Y=25x5000 signal(double)
3. [m,n0] = size(D0);
4. [~,N] = size(Y);
5. itcmat = zeros(iters/iter_check, n0-nmin+1);
6. D0run = D0;
7. ii = 1;
8. vn = [n0]; % vector of actual sizes
9. for it = 1 : iter_check : iters
10. % run the algorithm for a few iterations
11. [D, X, ~] = DL(Y, D0run, s, iter_check, str2func(update), params, 'replatom', replatom);
13. % order dictionary based on representation power
14. [~, atom_list] = sort(sum(X.*X,2), 'descend');
16. % compute ITC for all dictionary sizes (with ordered atoms)
17. n1 = max(nmin, vn(ii) - ncand);
18. for n2 = n1 : vn(ii)
19. Dn = D(:, atom_list(1:n2));
20. Xn = omp(Y, Dn, s);
21. rmse = norm(Y-Dn*Xn, 'fro')/sqrt(m*N);
22. if itc_index > 10
23. itc2 = com_itc_BS_2( Y, Dn, Xn, rmse );
24. itcmat(ii,n2-n1+1) = itc2(itc_index-10);
25. else
26. [~, itc2] = com_itc_BS( Y, Dn, Xn, rmse );
27. itcmat(ii,n2-n1+1) = itc2(itc_index);
28. end
29. end
31. % trim or enlarge dictionary
32. [~,n2] = min(itcmat(ii,1:vn(ii)-n1+1)); % compute min value of ITC
33. n2 = n2 + n1 - 1;
35. nopt(ii) = n2; %% error Subscripted assignment dimension mismatch.
37. itcmat(ii,vn(ii)-n1+2:n0-n1+1) = itcmat(ii,vn(ii)-n1+1); % fill with last value of ITC
39. if n2 < vn(ii) - nminus % dictionary is too big
40. n2 = max(nmin, vn(ii) - nminus);
41. D0run = D(:, atom_list(1:n2)); % continue with new dictionary
42. elseif n2 < vn(ii)-1
43. n2 = max(nmin, vn(ii) - 1);
44. D0run = D(:, atom_list(1:n2)); % continue with new dictionary
45. elseif n2 == vn(ii) % dictionay may be too small, increase size
46. %n = vn(ii);
47. n2 = vn(ii) + nplus;
48. Dplus = randn(m,nplus);
49. Dplus = normc(Dplus);
50. D0run = [D Dplus]; % continue with new dictionary
51. end
52. vn(ii+1) = n2;
54. % final operations
55. X = omp(Y, D0run, s); % compute RMSE just to see it
56. rmse = norm(Y-D0run*X, 'fro')/sqrt(m*N);
57. rmsev(ii) = rmse;
59. ii = ii+1;
60. end
62. % final refinement: a few DL iterations with the final size
63. if nopt(ii-1) < vn(ii-1)
64. D0run = D(:, atom_list(1:nopt(ii-1))); % cut dictionary to optimal number of atoms
65. end
66. [D, ~, ~] = DL(Y, D0run, s, iter_check, str2func(update), params, 'replatom', replatom);
67. X = omp(Y, D, s);
68. rmse = norm(Y-D*X, 'fro')/sqrt(m*N);
71. % final operations
72. X = omp(Y, D0run, s); % compute RMSE just to see it
73. rmse = norm(Y-D0run*X, 'fro')/sqrt(m*N);
74. rmsev(ii) = rmse;
76. ii = ii+1;
77. end
79. % final refinement: a few DL iterations with the final size
80. if nopt(ii-1) < vn(ii-1)
81. D0run = D(:, atom_list(1:nopt(ii-1))); % cut dictionary to optimal number of atoms
82. end
83. [D, ~, ~] = DL(Y, D0run, s, iter_check, str2func(update), params, 'replatom', replatom);
84. X = omp(Y, D, s);
85. rmse = norm(Y-D*X, 'fro')/sqrt(m*N);

[Adam]

You should be able to determine what the problem is in 10s using the debugger and the 'Pause on errors' option. It is a lot more difficult to just scan 85 lines of code and try to work out what is happening.

nopt is not predeclared, which may be the cause, but equally if n2 is non-scalar on any iteration of the loop then this would give this error, as it would also if n2 is empty ( [] ).

[Guillaume]

In my opinion, it would be much easier to read the code if it had been formatted as code (with no line numbers) using the button.

The error is because ii and n2 are not the same size. As Adam says, the debugger is the best way to find out what size they are.

We have

  [~,n2] = min(itcmat(ii,1:vn(ii)-n1+1));   % compute min value of ITC
  n2 = n2 + n1 - 1;
  nopt(ii) = n2; %% error Subscripted assignment dimension mismatch.

So, n2 (at least this n2 since n2 is used for many purposes including the loop index. An infinity of variable names available, yet n2 has to be reused many times!) is the index of the minimum of a row of a matrix. So it must be scalar.

Most likely n1 is not scalar. n1 is the max of two things which do not appear anywhere else in the code, so we don't know what their size is. Most likely, nmin or ncand is not scalar.