In determining the point spread function (PSF) of digital images, the estimation of the parameters of an ARMA model is usually reduced from a 2-D problem into a 1-D problem. Does anyone know of the necessary conditions for this to be possible? A reference could be very useful. I would also appreciate some comments about the fitting of spatial models to digital images corrupted with noise. Thanks Coomaren Coomaren P Vencatasawmy Scientist Swedish University of Agricultural sciences Forestry Department Umeå [log in to unmask] http://www.resgeom.slu.se/eng/default.htm %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%