Using Machine Learning to Estimate Some Anisotropy Indices, Application to Brownian Textures and Breast Images

In this paper, we analyze image textures with help of anisotropic fractional Brownian fields. We also use some anisotropy indices characterizing the anisotropy of these textures. Multi-oriented quadratic variations form the basis of mentioned indices. Anisotropy indices are invariant to some image transforms. Furthermore they can be estimated from the observed data. An application of these indices, combining with a measure of texture roughness, is in lesion detection in mammograms.