The study of regional spatial structure and growth patterns has been of interest to those concerned with regional planning and analysis. Spatial structure can be defined in various ways, and one major perspective relates to investigating the distribution of population densities. The population density function, which has been applied widely within the context of an urban area, can be extended to analyzing regional spatial structure and growth patterns. However, most of the empirical studies so far conducted have employed the density function based on the monocentric model, which is not appropriate for modeling the modern metropolitan region that usually takes a polycentric form. This study first discusses the form of polycentric regional density function, and then applies it to the Beijing-Tianjin-Hebei metropolitan region. The polycentric density function is assumed to be an aggregate of the individual monocentric density functions of the various regional urban centers. Through the exploratory spatial data analysis of the population density distribution, we find that the Beijing-Tianjin-Hebei metropolitan region does have a polycentric spatial structure with four regional density centers: Beijing, Tianjin, Baoding and Shijiazhuang. We test the most commonly used monocentric regional density function form, the square root negative exponential density function, and the results show that it is basically applicable to our analysis. Then we apply both the monocentric and polycentric density functions to study the spatial structure and growth patterns of the Beijing-Tianjin-Hebei metropolitan region during the period from 1990 to 2000. The analysis with the monocentric density function shows population still concentrate into the major urban centers of the region, while the analysis with the polycentric density function finds the coexistence of concentration and dispersion of population in the region, and reveals three different growth patterns of the urban centers at the different levels, which are the dispersion pattern through decentralization, and growth and the concentration pattern. The study shows the polycentric density function, taking the effects of the multi-urban centers on the regional population distribution into consideration, can reveal the regional growth patterns more accurately and is more suitable to the analysis for the polycentric urban region.