Usually, 3D building reconstruction is realized by systems, which either apply constructive solid geometry (CSG) or boundary representation (B-Rep) to model the respective buildings. We present an alternative approach based on cell decomposition, which can be used efficiently for building reconstruction at different scales. Firstly, building polyhedrons are constructed from airborne LIDAR data and given outlines of the respective buildings. In this context, cell decomposition is used to guarantee topological correct representations and the simple implementa-tion of constraints between different building parts like co-planarity or right angles. As it will be demonstrated in the second part of the paper, cell decomposition is also advantageous if large scale building models have to be generated for example based on the analysis of terrestrial LIDAR data.