Time series model for texture synthesis

A general method is proposed based on a model which treats the pixels of a digitized textural scene as a two-way seasonal time series and the parameters needed for synthesis are derived directly from the analysis of the “parent” texture.

Abstract

A general method is proposed for the synthesis of texture. It is based on a model which treats the pixels (picture elements) of a digitized textural scene as a two-way seasonal time series. This method possesses the desirable characteristic that the parameters needed for synthesis are derived directly from the analysis of the “parent” texture (texture to be imitated). With the help of well-developed methods in the time series analysis the process that generates the pixels of the parent texture is identified. From a set of boundary conditions the future values of the time series are generated which in essence are the pixels of the synthesised texture. The effectiveness of this method is illustrated with examples.