Interpolation with interp1 and matrix

Hello,

I have a signal of 25,000 points recorded for several pixels (1,600). Each of the pixels have an associated grid of 25,000 points. I would like to perform a "pchip" interpolation of my signal to a new common grid for all pixels with a length of 20,000 and without a for loop. I have tried with interp1 (common grid and signal with dimensions 1,600 x 25,000; common grid with dimension 1 x 20,000) but it does not work without a for loop because to each pixel there is a different associated grid. I believe the approach would work if the original grid was common for all pixels.

Could you help me with that? Is there an alternative without a for loop?

Thanks in advance,

Francesc

[KSSV]

Whats wrong in using loop?

[Francesc]

I already have several loops on top, as I am analysing massive amounts of data. Therefore, I need to speed up every bit of code I have.

Answer 1

[Matt J]

%example data
x=sort(rand(25000,1600),1); %grids per pixel
v=rand(size(x)); %signals
xq=sort(rand(20000,1)); %common grid;

%loop method
tic;
xnew=nan(20000,1600);
for i=1:size(x,2)
    xnew(:,i)=interp1(x(:,i),v(:,i),xq,'pchip');    
end
toc

Elapsed time is 1.937708 seconds.

%loop-free method

tic;
increm=cumsum( max(x(:,1:end-1),[],1)+1 );
x(:,2:end)=x(:,2:end)+increm;
xq=[xq,xq+increm];

xnew=interp1(x(:),v(:),xq(:),'pchip');
toc;

Elapsed time is 3.196412 seconds.