In order to study the relationship among different vegetation communities, soil structure and infiltration characteristics in loess hilly region, and to quantitatively explore the soil structure, fractal theory and methods were used to study the relationship between soil fractal dimension and soil texture, bulk density, porosity, moisture content and saturated hydraulic conductivity of seven vegetation communities, i. e. , Pinus tabuliformis pure forest, Robinia pseudoacacia pure forest, P. tabuliformis + R. pseudoacacia mixed forest, Platycladus orientalis pure forest, Rosa xanthina shrubbery pure forest, Caragana microphylla shrubbery pure forest, and wasteland in Loess Hilly Region, Western Shanxi Province. The results showed that: 1) All kinds of vegetation communities were beneficial to improving the structure of the soil particles. The improvement of coniferous and broad-leaved mixed forest (P. tabuliformis + R. pseudoacacia mixed forest) is the best, followed by broad-leaved forest ( R. pseudoacacia pure forest and Rosa xanthina shrubbery pure forest), and coniferous forest ( P. tabuliformis pure forest and Platycladus orientalis pure forest) is the lowest. 2) The soil fractal dimension in the study area was between 2.799 and 2.805, and the mass fraction of clay ( < 0.002 mm) was inhigh level. 3) The soil fractal dimension had a significantly positive correlation with the mass fraction of soil clay, while a significantly positive correlation with the mass fraction of soil sand, total soil porosity, moisture content, and saturated hydraulic conductivity. Therefore, the soil structure and infiltration characteristics can be characterized by the fractal dimension of soil particles. The improvement effects of different vegetation communities on soil varied, and the results could provide a reference for the afforestation in the study area.