Since the 90’s, geostatistics are increasingly and better integrated in the earth model building process [1]. Geostatistics provide a suitable language and dedicated tools to analyse the spatial coherency of spatial data sets (mono-variate or multi-variate) and therefore lead to more robust interpolations, spatial filters and simulations. More precisely, the factorial kriging (FK) model, a geostatistical filtering technique developed in 1982 by Matheron [2], has been successfully used in seismic processing [3], [4], [5], [6], [7]. Nevertheless, this technique appears to be limited by structural dipping effects when applied to seismic amplitude or related attribute cubes.