You don't specify which dimension hold the matrices you want to get the minimum values of. However, you should be able to get the idea for an algorithm from this. You just have to properly reshape (and possibly shift some dimension around using permute) to line up the values you want the minimum of.
% Some pretend data
M = rand(2,3,4,5)
% Supposing I don't know the size, I get it
S = size(M)
% Reshape such that each array is now a column
R = reshape(M,prod(S(1:3)),[]);
% Get the minimum values
V = min(R)
