Image enhancement and microstructure characterization of energy dispersive X-ray spectroscopy images of blended cement pastes

Abstract

Energy dispersive X-ray spectroscopy (EDS) image analysis has been widely used to study the microstructure of cementitious materials, whereas in some cases the efficiency of the characterization suffered from the extremely long time for image acquisition. In this study, we proposed an EDS image analysis methodology including image enhancement, phase extraction, and characterization of reaction products. The guided filter and Logarithm correction were combined and employed to improve the resolvability of low-quality EDS images. Gaussian Mixture Model (GMM) algorithm was applied to segment the unreacted phases and two different groups of reaction products from the enhanced EDS images. Moreover, the spatial distribution and chemical composition of reaction products were characterized using the lineal-path function and simple lineal iterative clustering (SLIC) algorithm. Results showed that low-quality EDS images after enhancing could achieve the comparable performance with high-quality EDS images in terms of the phase proportions. This study offers a practical and effective methodology for enhancing and analyzing the low-quality EDS images, giving insights into the phase assemblage and microstructure evolution of cementitious materials.