how to find sub matrixes in large one

I try to find sub matrixes in large one. I read documents a lot, but didnt find functions support for this one please help me

[Stephen23]

Basically loop over the indices and check the match using your chosen difference metric. There is no magic solution.

[Image Analyst]

Well it might not be magic, but it might cut down on the number of lines if you use normxcorr2() and look for the max value, as I show in my demo below.

Answer 1

[Image Analyst]

If you have the Image Processing Toolbox, you can use normxcorr2(). Demo m-file is attached below the image.

[xuan hau nguyen]

I mean, I want to find areas (or sub matrixex) in the large matrix, this areas is separate and each have the same value

[Image Analyst]

And what do you think I did? I did exactly that. I defined a template to search for (a small matrix that I know for a fact exists somewhere in the larger matrix), and I found it in the larger matrix. If you don't like my demo, then supply me with matrices of your own and I will demonstrate that it will work with your matrices also.

[xuan hau nguyen]

what's you mean? I want ask about which function in matlab that support to find sub matrix (have same value or with conditional value) in large one. i think function normxcorr2 only suit for image.

[Image Analyst]

An image is a matrix, so normxcorr2() works for matrices. My offer still stands. If you post your matrices, I'll show you how it works.

[xuan hau nguyen]

this is my matrix, how find sub matrices which have the same value

[Image Analyst]

I'm sure you've already tried it and proven it to yourself by now, but for what it's worth, here is my demo with your data:

% Read in data
data = csvread('data.txt');
subplot(2, 1, 1);
imshow(data, []);
[rows, columns] = size(data);
% Data is 28 x 721

% Create a 8 by 4 template
template = [...     % Will be found at data(6:13, 3:6)
     0    52    52     0
     0     0    52     0
     0     0     0    52
     0    52    52     0
    52    52     1     0
     0    52     0     0
     0     0    75     0
     0     0     0     0]
[rowsT, columnsT] = size(template);

% Now we have our large matrix and our template and we can begin:
% Find the normalized cross correlation.
c = normxcorr2(template, data); % c is 35 by 724
subplot(2, 1, 2);
imshow(c, []);
% Extract middle 28 x 721
c = c(rowsT:rowsT+rows-1, columnsT:columnsT+columns-1);
% Find where template occurs in data
[matchingRows, matchingColumns] = find(abs(c - 1) < 0.00001)
message = sprintf('That pattern shows up at row=%d, column = %d\nof the larger matrix', matchingRows, matchingColumns);
uiwait(helpdlg(message));

[xuan hau nguyen]

I already a test for normxcorr2 function example

a =

       1     2     3     4
       1     2     3     7
       1     2     3     6

>> b

b =

       1     2
       1     2

>> normxcorr2(b,a)

ans =

      0.5774    0.3015    0.1925    0.1400   -0.5774
      1.0000    1.0000    1.0000    0.7625   -0.9330
      1.0000    1.0000    1.0000    0.9802   -0.9941
      0.5774    0.3015    0.1925    0.3015   -0.5774

I saw the result have multiple location of matrix b but really b only two in matrix a. how are you know which one exactly for matrix b in matrix a.

[Image Analyst]

Well why didn't you just do like I did? matchingRows and matchingColumns will find both locations. I had two extra lines of code after normxcorr2(). Why did you not use those? I didn't put them in for nothing. You were supposed to use them!

[xuan hau nguyen]

thank you so much !!!!!!

[xuan hau nguyen]

but when i find Template that all of item the template have the same value then happen a error Error using normxcorr2>checkIfFlat (line 283) The values of TEMPLATE cannot all be the same.

what I must do this case?

[Image Analyst]

What was your template?

[xuan hau nguyen]

example a my Tamplate [1 1 1 1 1 1; 1 1 1 1 1 1; 1 1 1 1 1 1]