Finding an average matrix from each matrix in an array

I have an array of matrices, each with the same dimensions, I am looking to get one matrix which has the average value of all the matrices at each point, any advice on how to do this?

[Joseph Cheng]

i think the matlab coding help would depend on how you've arranged your array of matrices. but over all if you have matrix of N dimensions and you want the index average across the different matrices; the simplest would be to just add all these together to a single N dimensioned matrix and divide that by total number of matrices.

Answer 1

[Image Analyst]

averageMatrix = (matrix1 + matrix2 + matrix3) / 3;

[Roisin Coveney]

Thanks, Its just I will have say 1000 matrices in each array so didn't want to write out each one, can I use a for loop for that?

[Image Analyst]

Yes. If you have a cell array, ca, with 1000 cells

numberOfCells = numel(ca)  % works for either 1D or 2D or more cell arrays.
sumMatrix = zeros(size(ca{1})); % Initialize to all zeros.
for k = 1 : numberOfCells
    sumMatrix = sumMatrix + ca{k};
end
% Compute the average from the mean.
averageMatrix = sumMatrix / numberOfCells;
imshow(averageMatrix, []); % Display the average matrix in an axes control.

Note that the matrices in all cells must be the same size, otherwise you can't add them.

But a big question is why the matrices ended up in a cell array instead of a regular numerical array, like a 3-D double array. Cell arrays are slow, inefficient, memory hogs. I think a structure array or table would also be less efficient than a 3-D double array.

https://matlab.fandom.com/wiki/FAQ#What_is_a_cell_array.3F

[Roisin Coveney]

Thank you! My data isn't in a cell array it is in a structural array, when I use that code the following message comes up

Brace indexing is not supported for variables of this type.

[Roisin Coveney]

Thanks, just changed it a bit and thats working now!

Answer 2

[Jan]

Which kind of array contains the matrices? A cell array, a numerical 3D array, a struct array? The solution depends on this detail.

A = {rand(4), rand(4), rand(4)};  % Cell array
meanA = sum(cat(3, A{:}), 3);

B = rand(3, 4, 4);   % 3 4x4 matrices in numerical 3D array
meanB = reshape(sum(B, 1), 4, 4);

C(1).data = rand(4);  % Struct array
C(2).data = rand(4);
C(3).data = rand(4);
meanC     = sum(cat(3, C.data), 3)

[Roisin Coveney]

the data is contained in a structural array