Parameter estimation for two-dimensional ising fields corrupted by noise

On a finite sublattice [Lambda] of Z2, consider a free-boundary Ising model X with inverse temperature [beta] > 0 and without external field. Assume that each spin is flipped with unknown probability [epsilon], independently for each site and of X. This has been suggested as a stochastic model for digital images. In this paper estimators are proposed for [beta] and [epsilon] and shown to be consistent as [Lambda][short up arrow]Z2. They are very easily computable since they do not require any evaluation of conditional statistics. Numerical experiments are reported on the performance of these statistics on a lattice of moderate size.