A spatialized model of visual texture perception using the structure tensor formalism (Q1953030)

Based on two previous works with \textit{O. Faugeras} [J. Nonlinear Sci. 21, No. 4, 465--498 (2011; Zbl 1229.35014); J. Math. Neurosci. 1, Paper No. 4 (2011; Zbl 1259.92007)] the authors analyze a spatialized network of interacting hypercolumns in the context of texture perception in the primary visual cortex, described by Wilson-Cowan neural field equations set on an abstract cortex \(\mathbb{R}^2 \times \mathrm{SPD}(2,\mathbb{R})\). Here \(\mathrm{SPD}(2,\mathbb{R})\) is the set of structure tensors. The coupling function for this network is characterized by local and long-range connections and can be written explicitly. Two previous articles allow to consider only structure tensors with determinant equal to one and identify the considered space with Poincaré disk \(D\). Using the symmetry properties of these connections the authors investigate the spontaneous patterns formation of cortical activity. Methods of equivariant bifurcation theory allow to determine selection and stability of patterns.