Can I smooth 1D data taken from pages of 3D array without using for loops?

I am looking to take a 3D array (~1000x1000x18) and take the vector along each page at each x,y point [e.g. (1,1,:); (1,2,:) etc.] and smooth or fit it to find the max value at each point. Each of the vectors are noisy cosine squared and I want to be able to get a true max at each value. Right now the noise is such that max(X,[],3) gives a different result than I want. So I want to smooth or fit each vector of data to get a better max of each vector out.

I can think of a brute force way to do that using for loops to pull each vector out one at a time but was wondering if this a nicer way.

[Matt J]

Do the cosines have different peak locations, or just different amplitudes?

Answer 1

[Matt J]

You could apply a 1D convolution along the 3rd dimension as follows,

 kernel=reshape(kernel,1,1,[]);
 convn(yourarray,kernel,'same')

[Matt J]

FFT low pass filtering processing might be even better, since your square cosine is highly bandlimited

F=fft(yourarray,[],3).*lowpass_mask;
f=ifft(F,[],3);

   result = max(f,[],3);

[Ben C]

Ok, that looks promising, thanks