Abstract: Scanning electron microscopy(i.e. SEM) is a commonly-used technique for qualitatively analyzing the soil microstructure. It could provide two-dimensional quantitative information on grain-size or pore-size distribution of the soil combining with image processing techniques. However, typically when this method is applied, grey binarization technique is used to distinguish soil particles from pores in a micrograph, which would result in great errors. In the present study, an image processing technique using Photoshop and MATLAB software in combination is proposed, for quantitative analysis of the microstructural characteristics of loess. The proposed technique is used to process the micrographs before and after intact and saturated Malan loess consolidation. The variations of pore-size distribution characteristics due to loading and wetting are interpreted and the influence of microstructure on loess collapsing is understood. The results indicate that the proposed image processing technique in combination with SEM can be used for accurate quantitative analysis on the loess microstructure and its evolution.