Impervious surfaces are directly related to urban development and population growth and they can be used as indicators of urban infrastructures. This work aimed to estimate urban population from impervious surface modeling considering two Brazilian cities of similar population size and different urban context: Sinop, in Brazilian Amazon, state of Mato Grosso, and Guaratinguetá, a city in the Southeast Region, state of São Paulo. A linear spectral mixing model was used to map impermeable surfaces from Landsat images. Using population density from 2010 Census, and impervious surface intensity, a linear model was proposed to estimate population. The absence of linear relations in Sinop indicates that other approaches to relate population and urban structures should be tested for Amazonian cities. In Guaratinguetá, significant R2 indicates that impervious surfaces is helpful to infer population density. Despite the high average relative errors for both study sites, the methodology demonstrated efficiency in the mapping of impervious surfaces.