Accurate knowledge of an imaging system's point spread function (PSF) is crucial for successful deconvolution. For fluorescence microscopy, PSF estimations based on either theoretical models or experimental measurements are available. However, due to the axially shift-variant nature of the PSF, neither method guarantees an estimate that is valid for the entire object space. In this work, we present a reduced-parameter version of a state-of-the-art theoretical model. We give a maximum-likelihood based algorithm for the estimation of its parameters, and we show how a fit of our model to two axially isolated point source measurements in an experimental setup can be used to accurately reproduce measured PSFs within the entire specimen.