Comparative Study on Backscattered Electron Image of Cement Paste Quantified by Frequency Histogram, Overflow and Gaussian Distribution Fitting Method

Abstract

Backscattered electron (BSE) images have been widely used to investigate the phases of cementitious materials. With the view of quantifying phases including pores, hydration products, and unreacted clinkers from the cementitious products, three quantification methods based on grey level thresholds have been proposed to segment phases from the BSE image. In this paper, these three phase-segmentation approaches, including frequency histogram method, overflow method, and Gaussian distribution method, are systematically studied and critically compared. Results show that the overflow method and the Gaussian distribution method outweigh the frequency histogram method in terms of human visual judgment, phases threshold, phases content, reconstructed image quality, and porosity. Furthermore, the overflow method seems to be more practical for phase quantification, producing comparable results to the Gaussian distribution method. The Gaussian distribution method possesses the potential for quantifying over three types of phases.