Density is probably one of the most used indicators to characterize urban development. Even if it is a quantitative and properly defined measure, there are still problems to use it properly. This paper proposes a renewed approach to characterize urban density based on buildings' footprints. It can be applied on huge datasets and allows multi-level characterization of density. We first present an original partition of urban open space. This topology helps us to define a neighborhood function. We then adapt the GSI and FSI indices to the previously defined tessellation. The combination of the neighborhood function and the modified indices makes it possible to assess density iteratively. For each building, these values allow to define density profile which is then used in a classification process. The results highlight spatial patterns and homogeneous areas. This transposable method is adapted to urban fabric characterization and surpasses old descriptive and low formalized classifications.