Geostatistical “filters” using Factorial Kriging are increasingly used for cleaning geophysical data sets from organized spatial noises that are difficult to get rid of by standard geophysical filtering. The understanding and handling of such kind of spatial processing is not easy for geophysicists who are neither used nor trained to handle stochastic models. In this paper we demonstrate the formal equivalence between Factorial Kriging models and usual geophysical filters (Wiener, (F,k), Median) and the added value of stochastic modelling that is the quantification of the quality of the filtering process. Cases studies on pre stack gathers and VSP wave separation illustrate the fact that these stochastic techniques are generic and apply to all filtering contexts.