The detection of spots with spot detector is based on wavelet analysis, that determine the "relevant" signal in the image. The spots are then segmented and their localization is obtained through their center of mass. The localization can be refined by using the center of intensity which is the intensity-weighted mean of spot pixel positions.
When using a Gaussian fit procedure, the localization uncertainty is not directly linked to the variance of the fitted Gaussian profile and depends on the number of emited protons, the wavelength of the dye and many other parameters.
I recommend you the following article that explores this localization issue: Raimund J. Ober,*† Sripad Ram,‡§ and E. Sally Ward†‡ ,Localization Accuracy in Single-Molecule Microscopy, Biophys J. Feb 2004; 86(2): 1185–1200.
In addition, it has been recently shown empirically that Gaussian fitting and wavelet analysis provide the same localization accuracy (Izeddin et al, Optics express, 2012)
Thibault Lagache