2-D optimized linear interpolation

naninterp2.m

The griddata Matlab's core function may fail with memory issue on huge grids, depending on your computer capabilities, even for simple linear interpolation to fill gaps. naninterp2 is a simple function that optimizes the linear interpolation by reducing the amount of relevant data (only gap neighbours) before calling griddata. This results in a very fast computing, as the speed mainly depends on the amount of gaps and less on the amount of valid data (i.e., size of the grid). See benchmarks below.

Example

z = peaks(1000);
% simulates random gaps of different sizes
for n = [10,5,2]
  z(ceil(rand(1e4/n^2,1)*numel(z))) = NaN;
  z = conv2(z,ones(n)/n,'same');
end

subplot(121)
imagesc(z); % with gaps
subplot(122)
imagesc(naninterp2(z)) % gaps filled

Benchmarks

This a comparison of computation times of interpolation for a 1000x1000 grid with about 5% of NaN contain (upper example) when using griddata and naninterp2 functions.

code	Matlab 2020a	GNU Octave 5.1
griddata.m	4.2 s	3 mn
naninterp2.m	0.16 s	12 s

Author

François Beauducel, IPGP, beaudu, [email protected]

Documentation

Type doc naninterp2 for help and syntax. See also users community comments.