Global and multi-scale image analysis using power spectra

Image analysis is currently used to study the structure of materials. In this paper, classical techniques of signal analysis such as power spectrum and autocorrelation functions are applied for this purpose. These techniques allow the detection of hidden features in complex images such as fracture surfaces, before highlighting them using spatial filtering. For that, the power spectrum allows us to locate the range of scales of peculiar texture, present but invisible to the naked eye. Thus, this study deals with the homogeneity of epoxy networks. Another way to use the method concerns aggregation phenomena. Indeed, the method allows us to assimilate several images of different magnifications, in order to characterize the aggregation on a large range of scales from the size of the particle or fibre to the full size of the sample. Then, the peculiar morphology of thermoplastic composites made from commingled fibres can be studied. The autocorrelation function method is also used to highlight features in the images of such composite materials.