Summary: Bayesian image classification methods, namely maximum marginal posterior estimation and the iterated conditional modes algorithm \(\psi\) are applied in the estimation of biogeographical ranges. The presented procedures are developed to utilize atlas map data, that is, observations recorded in a grid of cells, by interpreting these maps as incompletely observed binary pixel images. Modelling of breeding bird atlases with a complex presentation of observations and inhomogeneous research activity is especially considered. Empirical examples are provided.