Interpolating data in a better way?

I will simplify this down, as my data is huge. I have about 100 sets of data smaller than the one I am going up to; each one of them vary in size. IE 3x6 or 9x9. I need to interpolate it up to say 20x20. Names of variables were changed for simplicity also.

A 3x3 raw data would look like this, and cannot change (easily) as I am importing from an edit locked file.

rawdata=[1 7 3 ;8 9 6; 2 2 5];

Here is how am currently going about it.

g_raw = size(rawdata);

xg_g = linspace(1,20,g_raw(1));

yg_g = linspace(1,20,g_raw(2));

interp_x_g = interp1(xg_g, g_raw, 1:20); %doing one direction

interp_y_g = interp1(yg_g, interp_x_g', 1:20)'; %tackling other direction with a transpose and transpose back

interpolated_g = interp_y_g(:); %making into a vector for plotting vs another vector

It works in a sense, but comes out very choppy. Is there a better way? I have been using MatLab for about a week and getting the hang of it. I just do not understand everything in the help files. IE I could not get polyfit to work in my loops. Not sure if I should be using interp2 or 3? All help is appreciated, even if it is just correcting syntax or general advice.

Cheers!

Walt

Answer 1

[dpb]

Yeah, try interp2 -- sotoo

[nr,nc]=size(rawdata);
[X,Y] = meshgrid(1:nr,1:nc);   % the original array mesh
[Xq,Yq] = meshgrid(linspace(1,nr,20),linspace(1,nr,20));  % the 20X copy
Vq = interp2(X,Y,rawdat,Xq,Yq);

[Walter]

I got that working with some edits. Turns out if you do my method and then yours and plot them against each other you get a one to one relationship. So I guess interp2 is doing exactly the same thing but in a much cleaner manner. Thanks!

[dpb]

So yours was, indeed, correct! :) I didn't read closely enough to tell but yes the two are the same net result.

You mentioned roughness; the alternative would be a smoothing instead of interpolating spline if the beginning points don't have to be reproduced identically.