This small contribution is useful in cases where you search for the position of a peak within an intensity matrix (correlation or energy fields) of low resolution. Since this is only a 2nd-order-fit the sub-sample accuracy will only be acceptable as long as the peak area is 'smooth' around the peak location. This is not an exact solution, since it does not consider any underlying signal characteristics.
Sample usage: % 2d-gauss: center=19.5; sigma=5
>> M = exp(-((1:30) - 19.5).^2/(2*5^2));
>> P = peakfit2d(M'*M);