Geovisualization applications that allow the navigation between maps at different scales while zooming in and out often provide no smooth transition between the individual building level of abstraction and the representation of whole urban areas as polygons. In order to reduce the cognitive load of the user, we seek to add intermediate zoom levels with intermediate and progressive abstractions between buildings and urban areas. This paper proposes a method based on progressive block graying while enhancing building landmarks, to derive these intermediate representations from the individual buildings. Block graying is based on an automatic building classification, and a multiple criteria decision technique to infer inner city blocks. The landmarks identification relies on machine learning and several criteria based on geometry and spatial relations. The method is tested with real cartographic data between the 1:25k (with individual buildings) and the 1:100k scale (with urban areas): transitions with one, two, or three intermediate representations are derived and tested.