Optimised wold-like decomposition of 2D random fields

In this paper we address the problem of image texture modeling. In particular we adopt here a 2D Wold decomposition that separates a texture into regular and chaotic parts, allowing for simple texture parameterization, and for explicit extraction of periodic structures. The identification scheme based on the proposed decomposition improves the accuracy of the estimation of the model parameters as well as the visual resemblance of the model with respect to samples, owing to compliance of the chaotic component with the Julesz's conjecture.