Towards a statistical treatment of images acquired by automated mineralogy

A Mineral Liberation Analyzer (MLA) is a scanning electron microscope, delivering quantitative data on mineralogy and microfabric as well as high resolution spatially resolved mineralogical composition images of rock specimen and particle perpetrates (usually grain mounts). MLA images can be processed to extract 2D information, both compositional (modal mineral composition) and microstructural (mineral associations within a particle, average particle diameter distribution, average grain diameter distribution, etc.). Unfortunately, the available software does not fully take into account the limitations of the actual information available: data are collected over a surface while information would actually be required over a volume; and most of them are actually relative, i.e. compositional. Moreover, there is virtually no statistical method or software devised to process data sets of this kind. This contribution presents this kind of data, their particularities and problems, and provides avenues of discussion towards a solution to these issues.