An important task when describing images using mathematical models is the identification of parameters from an real image. Such a task is of particular difficulty for combined mathematical models, which, among other things, make it possible to describe multidimensional spatially inhomogeneous signals. An example of such a model is a doubly stochastic model based on autoregression with multiple roots of characteristic equations, which has an important property of quasi-isotropy and allows describing a smooth change in the statistical and correlation properties of simulated images. This work is aimed at developing algorithms for identifying the parameters of this model based on the estimated probabilistic properties of the available image.