Moving window in time domain

I have a 3D matrix X*Y*T and I would like to apply a bandpass filter in the 3rd dimension. Instead of doing it on the complete series at once, I would like to have a window of size n for example (X,Y,1:n) on which I filter and then move to the next window (X,Y,n+1:2n) and so on. What is the most efficient way of doing this?

Answer 1

[Image Analyst]

The convn() function.

n = 3;
kernel = repmat(1, 1, 1, n) / n;
smoothedImage = convn(image3d, kernel, 'same');

[ecartman22]

Thank you for the tip. Could you kindly help with how it would look like in code? This is what I am doing right now.

P = rand(600,400,1200);
Wn=[0.6 / 30 * 2, 2.5 / 30 * 2];
[b,a] = butter(6,Wn,'bandpass');
y = filter(b,a,P,[],3);

[Image Analyst]

See untested code I added above. It will average over n slices in the z dimension then scan all over to hit every (x,y) location. The image can have more slices than n. For example, you can have 50 slices and n can be 3 then it will average over 3 slices at a time, then move down a slice and repeat until all slices have been scanned.

[ecartman22]

Thanks a lot! This is averaging over a window. How could I change the kernel so it bandpass filters the window in the time domain?