Modeling geometric ob jects is inescapable in various domains and applications. The development of acquisition and storage tools has produced a huge increase of the volumes used to store these ob jects. Many disciplines aim to reduce the size of these volumes ; such as compression, which tries to compress the volume, and the elaboration of theoretical compact data structures that minimize the space required for the representation. The goal of this thesis is to design, and evaluate practical and workable solutions for compact representations of triangulations. To do this, two issues has to be explored : modify the internal memory representation of the geometric ob jects, and redefine the abstract types associated to the geometric ob jects. Th...