Speed up 1D spline interpolation on matrix

Hello,

I have a NxM matrix named data and I need to spline interpolate each row of my matrix according to:

for k=1:size(data,1)
    data(k,:)=spline( x   , data(k,:).',  x + shift(k,:).' );
end

where x is the original monotonously increasing spacing of each row. And the NxM matrix shift is the row- and column-dependent shift that I want to apply to each sample.

Is there a way to speed this up further? I found a lot of functions that claim a speed improvement of x10 for linear interpolation. Anything I can do for spline?

Thanks!

Answer 1

[Matt J]

It might help to pose it as a 2D interpolation,

[m,n]=size(data);
idx=1:n;

F=griddedInterpolant({x,idx},data,'cubic');

uu= x+shift.';
vv=repmat(idx,m,1);

data=reshape( F([uu(:),vv(:)]),  m,n) ;