In view of a tentative investigation on the porosity analysis method of soil mesoscopic structure, and on the basis of Matlab image processing technology, by adopting the threshold segmentation and gray image recognition methods, a processing and analysis have been conducted on the SEM images of cement soil, thus obtaining the process map of cement soil microstructure and gray image recognition undergoing different threshold segmentation, followed by a comparative analysis of the images processed by the two methods. The results show that, for the same scanning area, the SEM images obtained under different multiple times can first determine the macroscopic surface porosity and find a reasonable threshold interval with the threshold segmentation method adopted. The threshold segmentation method is also applicable to the SEM image processing of specimens with relatively smooth surface. Meanwhile the gray scale image recognition method is suitable to the process of SEM image processing with obvious structural direction and single structural elements.