Interpolation 9 data. help!

Data is figure.

I want to interpolation "NaN" data.

Please help me.

[Jan]

Columnwise, rowwise, in 2D, linear, cubic or how else?

Answer 1

[Gavriel Aminov]

Please try the next code. The sample matrix M is simulated as random values, not as well-behaving function, therefore the ERROR is big. For smooth 2D function the ERROR should be reasonable. Please examine it.

m=5;
n=10;
M0=round(rand(m,n)*10)

M=M0;
M(M<3)=nan

[J, I]=meshgrid(1:n, 1:m);
all_inds=sub2ind([m,n], I, J);
good_inds=find(~isnan(M(all_inds)));
[I0, J0]=ind2sub([m,n], good_inds);

MM=round(griddata(I0, J0, M(good_inds), I, J, 'linear'))

ERROR=(MM(isnan(M))-M0(isnan(M)))'

Answer 2

[Gavriel Aminov]

https://www.mathworks.com/mat Interpolates (& extrapolates) NaN elements in a 2d array. (nan, inpainting, replace) labcentral/fileexchange/4551-inpaint-nans