MATLAB multi-dimensional matrix access

Question

Happy new year SO members!

My first post for the new year, hoping to get some luck! :D

I have a 4D matrix X whose size is 100, 100, 100, 3. I extract its submatrix with X(51,:,51,:) and expect the result to be a 100x3 2D matrix. But NO, the result is a 4D matrix with size=1, 100, 1, 3. How come??

An even more confusing result is:

I can use plot(X(51,:,51,i)) and plot(X(:,51,51,i)) with i=1, 2, 3 just fine
can't use plot(X(51,51,:,i)) with the same i

in short, MATLAB submatrix extraction won't result in a size reduced matrix? and the different sub-access behaves differently, namely, more careful with the last dimension?

Thank you!

Edit01:

For convenience, I'd show my test results with singleton and squeeze here: sub matrix access

use of squeeze:

squeezed

tmpearce tmpearce · Accepted Answer · 2013-01-01T05:17:06

As you've noticed, when you index into your 100 by 100 by 100 by 3 matrix with (51, :, 51, :) you get back a result of size [1 100 1 3]. The dimensions with size = 1 are called singleton dimensions.

MATLAB won't remove non-trailing singleton dimensions automatically (for greater than 2D matrices) - use squeeze to eliminate these dimensions. If the last dimension is singleton, this is handled automatically.

So in your case, you have

X = zeros(100,100,100,3);
Y = X(51,:,51,:); #% size(Y) = [1 100 1 3];
Y2 = squeeze(X(51,:,51,:)); #% size(Y2) = [100 3] - singletons removed

P1 = X(51,:,51,1); #% size(P1) = [1 100 1 1] which becomes [1 100] (2D)
P2 = X(51,51,:,1); #% size(P2) = [1 1 100 1] which becomes [1 1 100] (3D)

MATLAB multi-dimensional matrix access

1 Answers