How to apply arrayfun on a 3D matrix?

Hello,

I have the following matrix:

size(A) = [64 64 1000];

I'd like to apply the following function on each frame, along the 3rd dimension.

lasca_func = @(x)(mean2(x)/std2(x));

How can I do this via arrayfun, and not through for-loop?

Thanks in advance!

[Adam]

I don't think arrayfun can be used for that - it takes the array values in turn as linear indices so the structure of the array is not preserved and the function will be applied per element for the whole 64*64*1000.

Answer 1

[Stephen23]

"How can I do this via arrayfun, and not through for-loop?"

You can't, because arrayfun operates on each element of an array. You cannot tell it to operate along specific dimensions of an array. An alternative would be to split the array into a cell array and then call cellfun, but this would be slower than using a preallocated loop.

Unless you have a gpuArray the simplest solution is to use mean and std, and specify the dimension:

>> A = rand(64,64,1000);
>> Z = mean(A,3)./std(A,0,3);
>> size(Z)
ans =
   64   64

[Adam]

As I understood the question the result should be a 1x1000 result though with the mean/std calculation over each 64*64 matrix.

I was thinking something similar to this, but while it works for mean you can't do this for standard deviation as it doesn't give correct results:

mean( mean( a, 1 ), 2 ) ./ std( std( a, [], 1 ), [], 2 )

[Mark Golberg]

Thanks guys.