Abstract: The quantitative analysis of soil desiccation crack network is significant to the investigation of crack mechanisms and the evaluation of engineering properties of cracked soils. In this study, desiccation tests are conducted on three different soils under temperature controlled conditions, and the obtained desiccation crack patterns are subjected to several pre-processing operations, e.g., binarization, noise elimination and skelotonization, by employing the image processing technique. A series of parameters are proposed and measured to describe the geometrical and morphological characteristics of the desiccation crack pattern. Based on the statistical method, a probability density function is introduced to quantify the distribution characteristics of some important crack parameters. The results show that the desiccation crack pattern is significantly influenced by physical properties of soils. The surface crack ratio, number of nodes, total crack length, average crack width and number of clods generally increase with the increasing clay content and plastic index. The physical meaning of quantitative parameters show some differences but linked to each other. For example, there is strong and positive correlation between the surface crack ratio and the average crack width; the number of crack segments is conditioned by the number of nodes; the average area of clods decreases with their increasing number; and the fractal dimension of desiccation crack network increases with the increasing surface crack ratio. For a general crack pattern, the number of nodes per unit area, number of clods per unit area and surface crack ratio can be recommended as the three basic parameters for the quantitative analysis. The results of statistical analysis indicate that the average values of the crack length, width and clod area are strongly related to their distribution ranges as well as the most probable values determined by the probability density functions.