The presented work is motivated by a pharmaceutical issue. Indeed, the production of drugs involves a crystallization process in solution depending on several parameters and their quality is closely linked to its geometrical characteristics (size, shape, quantity ...). Therefore, it is necessary to control the distribution of the “crystal geometry” during the process. To meet this goal, an in-situ optical acquisition system provides images of the crystal population crystals’ population, giving access to the projected particles during the process. These images highlight the overlapping of the crystals (caused by the 2-D projection) and it is consequently difficult to individualize the particles for further characterization. Therefore, after a binarization process [1], the objective of this work is to investigate stochastic geometrical models so as to represent these binary images and to get the geometrical characteristics of the crystals.