Smoothing 3D Matrix in z-direction

Hey guys, I've been trying to find a solution for smoothing a 3D (512x101x1200) matrix in the z direction, but was unsure how to proceed

I have conventional used 2D data sets for this kind of smoothing (arbritarily I use something like a 5point smoothing) but I am having trouble adapting this now to the third dimension

So ideally I would proccess the matrix by smoothing it in the z-direction, if anyone can help me with this it would be greatly appreicated. Apologies I can't attach my actual data as it is too large but any 3D matrix is sufficient!

Thank you!

[Rik]

Have you considered convolutions? They are generally relatively easy to extend to higher dimensions once you have them working in lower dimensions. That is the reason my region growing function supports dimensions higher than 3, even if I don't think anyone will ever use that.

[Eunan McShane]

Hey, thanks for getting in touch. I hadn't before now, are you suggesting that by using your function I can convolve a series of neigbouring points into one?

Answer 1

[Ameer Hamza]

See convn(): https://www.mathworks.com/help/matlab/ref/convn.html. You can directly do 3D convolution.

Answer 2

[Markus]

Hi,

this works, but it's a slow solution:

Assuming you have a matrix called 'data' that is (512x101x1200)

for x = 1:512
    for y = 1:101
        smooth(data(x,y,:))
    end
end

For more details to adjust the span e.g. see: https://www.mathworks.com/help/curvefit/smooth.html#mw_62e54d1e-a0d3-4ccb-9dd8-142962c488b3

It's a nested for loop so it's not very fast. If you concatenated all z columns their beginnings and ends get smoothed together, hence you need to smooth them one by one.