Opposite of Interpolation: reduce size from a very detailed 2D matrix to a less detailed 2D matrix

Hello,

If you look at this link:

https://de.mathworks.com/help/matlab/ref/interp2.html

You see the "Original Sample" interpolated to the linear interpolation grid. I would like to do just that but exactly opposite. My matrices are 2D but very big files (20GB). I would like to reduce the size but be able to still keep the form of the plots just like in the original sample. Meaning, save bytes but be able to still make some reasonable conclusions about the data.

Anyone have any ideas?

Answer 1

[Star Strider]

You can always interpolate to a coarser grid than the original data.

This uses interp1 for simplicity. The same idea will work for interp2:

x = linspace(0, 10, 100);
y = sin(x * 2);

xi = linspace(0, 10, 33);
yi = interp1(x, y, xi, 'linear','extrap');

figure(1)
plot(x, y, '-b')
hold on
plot(xi, yi, 'pg')
hold off

Answer 2

[dpb]

W/O knowing anything at all about what the data are it's tough to say anything specific but the most obvious is to simply use something like

x=x(1:n:end,1:n:end);

where n=1,2,3,... to decimate. After that, see <smoothing-data>