3 D Furlike Texture Generation by a 2 Dautoregressive Synthesis

This paper presents a texture model that allows a two dimensional synthesis of 3D furlike texture. The model is based on the modelling of the texture autocorrelation function (ACF) and a two dimensional AutoRegressive synthesis (2D AR). We consider that the appearance of a furlike texture is deened by two characteristics: the orientation and the length of each individual element (i.e. lament). The values of these characteristics are constant in a stationary texture. The AR synthesis model of a given texture is deened by a number of parameters which can be computed from the texture ACF. Therefore, we have built a mathematical model of the ACF with respect to orientation and length. The model obtained consists of an analytical expression of ACF with respect to orientation. For the length, we have used a lookup table. Indeed, the ACF expression includes some parameters which are numerically estimated for several values of length. This allows to build a lookup table indexed by length and containing at each entry a set of ACF parameters. In this way, a furlike texture synthesis based on this table can be implemented. We show that with an appropriate choice of other parameters of the AR synthesis, this approach yields satisfying results, essentially for the simulation of fur appearance on 3D surfaces.