Shifting center after resampling real array

[hzarei4]

Hello,

I wonder why the array's center is shifted after resampling to a larger size...

for example:

julia> a = disc(Float32, (11, 11), 0.2);

julia> argmax(a)
CartesianIndex(6, 6)

julia> b = resample(a, (111, 111));

julia> argmax(b)
CartesianIndex(51, 51)

The peak of the resampled array is at (51, 51) and not:

julia> size(b) .÷2
(55, 55)

[roflmaostc]

The reason is that we do resample around the first pixel (1,1, ...).

Try fftshift(resample(ifftshift(a), (111, 111))));

[roflmaostc]

Maybe you could add a simple example to this documentation page?

https://nanoimaging.de/FourierTools.jl/dev/resampling/

[hzarei4]

Thanks! It solved the problem.
Ok, I will add this to the documentation page. Or I can include this in the resample function (?)

[roflmaostc]

We could add it as a flag shift_center=false and then exactly doing those changes.
If you could do a PR, I can review :)

[RainerHeintzmann]

I agree. Such a flag would really be useful. However, we may think about how to implement this best. Two extra fftshift sound unnecessarily complex.

[roflmaostc]

On the other hand, the whole algorithm calculates in a O(n*log(n)) complexity whereas a fftshift is only O(n). I'd argue, we keep the default the same, and otherwise it's slighty more expensive but barely noticeable (only memory is a bit higher).

[RainerHeintzmann]

We can compare both options. Either a double shift or a separable multiplication in Fourier space.