Filter out NaNs but not lose data completely?

Hello!

I have some 3d data (x, y, z coordinates in 3 columns; x and y are physical coordinates, z is the result of a measurement) from multiple scans, with these I would like to create an averaged image (averaged in z, the data in x and y is always the same for all scans, i.e. the resolution matches). The individual scans have sporadic NaNs in the z dimension. So far I had created some averaged images by simply adding all the z data together, filtering out the x, y and z rows where there are NaNs in z, finally dividing z with the number of scans that are to be averaged and putting all the columns back together.

However, because the NaNs in z are kind of random, now I encounter a problem that there are more NaNs than I expected, which then in the averaged image eats away too much of the data, especially since I average about 10 images or so.

Is there a way for me not to filter the NaNs out completely but to do something of the following kind:

If all z values at cetrain x and y are NaN → then delete x-y-z row as described above;

If from z values at certain x and y only some are NaN → keep row by using average of non-NaN z-values as the new, averaged z?

Best regards!

Answer 1

[KSSV]

You can fill those NaN's using multiple ways. Read about fillmissing. You can also use interpolation methods to fill those nanas. Read about scattredInteprolant.

[Marko D]

Fillmissing got the job done. Thanks!