It is one of important subjects in earth and planetary sciences to analyze the textures of rocks quantitatively. For the purpose some digital imaging techniques, including back-scattered electron imaging and characteristic X-ray imaging with a scanning electron microscope (SEM) or a electron probe micro-analyzer (EPMA), and optical imaging with a CCD digital camera or a image-scanner etc., have been applied. We applied here a scanning X-ray analytical microscope (SXAM) for the first time in the field of earth and planetary sciences to obtain XRF images of rocks. In this paper, we report the new method of image processing in which the X-ray maps of a rock are transformed to the maps that show distribution of mineral composition. As a test case, the X-ray maps of the Ryoke granite from Teshima, SW Japan, were processed to make the distribution maps of major minerals. XRF intensities of major elements were assumed to have linear relationship with the composition of major rock-forming minerals in each pixel. The coefficients between XRF intensities and mineral compositions were determined by picking up the some pixels at which a pure mineral exists. The composition of minerals was then calculated by maximum likelihood (ML) method for Gaussian distribution i.e. least-square method. It was shown that it was plausible by a numerical experiment to adopt the least-square method when the operation time of SXAM is sufficiently long. We found the sources of errors of the processed mineral maps depending on statistical errors of X-ray counts, variation of chemical composition in each mineral, and a condition of a sample's surface etc., and propose the way of estimating the errors. As an application, the mineral maps was applied to modal analysis of minerals.
