Urban fabric characterization is very useful in urban design, planning, modeling and simulation. It is traditionally considered as a descriptive task mainly based on visual inspection of urban plans. Cartographic databases and geographic information systems (GIS) capabilities make possible the analytical formalization of this issue. This paper proposes a renewed approach to characterize urban fabrics using buildings' footprints data. This characterization method handles both architectural form and urban open space morphology since urban space can be intuitively and simply divided into built-up areas (buildings) and non-built-up areas (open spaces). First, we propose to build a mesh of the open space (a morphologic tessellation) and then we formalize relevant urban morphology properties and translate them into a set of indicators (using some common-used indispensable indicators and proposing a new formulation or generalization of a few others). This first step produces a highly dimensional data set for each footprint characterizing both the building and its surrounding open space. This data set is then reduced and classified using a spatial clustering process, the self-organizing maps in this case. Our method only requires buildings' footprints as input data. It can be applied on huge datasets and is independent from urban contexts. The results show that the classification produced is more faithful to ground truth (highlighting the variety of urban morpho-logic structures) than traditional descriptive characterizations generally lacking open space properties.