Fractal estimation in a given frequency range. Application to smectite images

In this communication, an efficient fractal estimation is proposed in a given frequency range. It is based on the maximum likelihood Whittle approximation for the stationary increments of fractional Brownian motion. Its efficiency is first shown on synthetic data generated by the Cholesky method. Then it is applied to smectite deposits on films analyzed by atomic force microscope. It is shown that the fractal parameter measured in the low frequency region allows to separate 3 groups of smectite clay images, and enables to recover chemical properties of the material.