Least squares image texture analysis and synthesis

Trygve Randen and John Håkon Husøy
Rogaland University Center
Department of Electrical and Computer Engineering
P.O. Box 2557 Ullandhaug, N-4004 Stavanger, Norway

Working Papers from Rogaland University Center, Submitted

Full paper


Stochastic texture modelling is a useful tool for revealing the mechanisms underlying the generation of natural textures. This paper presents a description of the calculation of the stochastic texture parameters, when we assume an autoregressive model. The parameters are estimated with a least squares estimator. The paper also presents an approach to texture synthesis using the estimated parameters. Experiments and comparisons of the results from both a causal and a non-causal model are presented.