(in french)Soft tissies modeling has many clinical applications, from computer assisted (minimally invasive) surgery and surgical simulators (for training) to detection of tumors or prediction of aneyrysmal rupture. Indeed, changes in vessel wall or organ elasticity can be indicative of diseased tissue.The main challenge in this domain is to combine computationnal efficiency and accuracy of modeling. Finite elements methods, derived from continuum mechanics, give precise physical deformation, but they are often slower than discrete models, whose main drawback is the lack of physical realism.We base our research work on a discrete model to represent soft tissues. This model is based on a shape memory that creates forces to simulate elastic deformations. In this document we investigate the relationship between its parameters and the biomechanics parameters commonly used. We also present a methodology for estimating theses parameters from an image of strains, which can be applied to elastographic methods.