Non-stationary 2D estimation: An application to turbulence field analysis

A two-dimensional (2D) nonstationary (NS) autoregressive (AR) model is proposed for the characterization of non-homogeneous textural patterns in images. The approach taken in this paper leads to the extension of a fast estimation technique to the 2D nonstationary case. The procedure which is used for estimating the parameters of the space-varying model is described in detail. This procedure is based on a relation between several 2D autoregressive models and on AR vector model. Next we give some preliminary experimental results and we discuss the applicability of this approach to modelling the vorticity field of a turbulent flow.