indexing diagonals out of a 4d matrix

I would like to extract diag elemnts of a 4d matrix and put them in 2d matrix. Any help would be much appreciated.

[madhan ravi]

Illustration needed.

[Bjorn Gustavsson]

As far as I understand the meaning of diagonal of an n-dimensional array A it ought to be composed of elements A(1,1,...,1), A(2,2,...,2), ...,A(n,n,...,n). You seem to be looking for some other type of diagonal, but if you can do the above with a loop you can do whatever you need with two nested loops.

HTH

[Adam Danz]

One way to communicate your needs is to supply a simple example of an input and the expected output.

[John D'Errico]

Give an example, as the diagonal of a 4-d matrix has no definition. If anything, as Bjorn points out, the result would be ONE dimensional, a vector. So wanting it to become a 2-d matrix makes little sense. That means you need to be clear and specific as to your intent.

[Neekar Mohammed]

Thank you very much for your comments. I have a correlation data which is a 4D matrix, let's say g(x1,y1,x2,y2) I want to calculate the intensity which is a 2D matrix f(x,y)=g where x1=x2 and y1=y2. So yes I want it to become a 2D matrix. Many thanks. Neekar

Answer 1

[Matt J]

[M1,N1,M2,N2]=size(g); 

 M=min(M1,M2); N=min(N1,N2);

f=diag(reshape(g(1:M,1:N,1:M,1:N),M*N,[]));

f=reshape(f,M,N);

[Matt J]

Nekar's comment moved here:

Thank you very much for fast respond. If is like g(x1,x2,y1,y2) (I changed the order) then f(x,y)=g for x1=x2 and y1=y2, Would be the same answer you suggested?

[Matt J]

No, you would have to pre-permute g into (x1,y1,x2,y2) ordering for it to work

g=permute(g,[1,3,2,4]);

For this reason and others, I recommend that you do not use (x1,x2,y1,y2) ordering. It will necessitate a lot of extra manipulation.