Vectorization instead of for loop

Hi.

I have an image with about 5000 objects on that I have the centroid (xf,yf) locations for. My aim is to take each one and perform a Gaussian Fit to. I currently use a for - loop and want to see if vectorization speeds it up, but I can't figure out how to.

heres my for - loop code:

l=numel(xf)     
delta=5;        %half width span of data to perform fit to
  for indx=1:l          
      xrange=xf(indx)-delta:xf(indx)+delta;     % create x range    
      ydata=B(yf(indx),xrange)';                %B=Original Image, so y is the intensity at the xrange positions
      xdata=(1:2*delta+1)';

        %Now do Gaussian fit      
                    [a(indx),b(indx),c(indx),d(indx),xpeak(indx),ypeak(indx),r2(indx)]=myGaussianFit(double(xdata),double(ydata), b0,c0);
           fwhm(indx) = c(indx) * sqrt(log(256));
           fwhmSUM=fwhmSUM+fwhm(indx);
           data(indx,1)=xf(indx);
           data(indx,2)=yf(indx);
           data(indx,3)=fwhm(indx);
           data(indx,4)=r2(indx);
           data(indx,5)=a(indx);
           data(indx,6)=b(indx);
           data(indx,7)=d(indx);
           data(indx,8)=xpeak(indx);
           data(indx,9)=ypeak(indx);
    end

Thanks Jason

Answer 1

[Sean de Wolski]

Preallocate your data array before the loop.

data = zeros(l,9);

Then populate it directly in the for-loop. The preallocation will be what speeds this up, not vectorization.

[data(indx,1),data(indx,2),data(indx,3),FIXME data(indx,etc)]=myGaussianFit(double(xdata),double(ydata), b0,c0);

[Jason]

Thankyou