Fast maximum likelihood algorithm for localization of fluorescent molecules

A common task in microscopy is to fit an image of a fluorescent probe to a point spread function (PSF) in order to estimate the position of the probe. The PSF is often approximated as a Gaussian for mathematical simplicity. We show that the separable property of the Gaussian PSF enables a reduction of computational time from O(L2) to O(L), where L is the width (in pixels) of the image. When tested on realistic simulated data, our algorithm is able to localize the probes with precision close to the Cramér-Rao lower bound.