For large datasets this allows much faster plotting.
An irregular 2D spatial sampling is provided in x and y vectors with a data point to plot in vector d. A bin size is also provded in vector b=[dx,dy]
Return a 2D matrix M with minimal empty cells with the correct spatial relationship of d and last value only retained in the event of duplicates.
b = [6,9];
x = rand(3333,1)*100*b(1)+1500;
y = rand(3333,1)*100*b(2)+50000;
d = sind(sqrt(x.^2+y.^2))./sqrt(x.^2+y.^2);
Return G, a 2D matrix with d binned in rows and columns