This article describes a process to include in a volumetric model various anatomical and mechanical information provided by different sources. Three stages are described, namely a mesh construction, the non-rigid deformation of the tetrahedral mesh into various volumetric images, and the rasterization procedure allowing the transfer of properties from a voxel grid to a tetrahedral mesh. The method is experimented on various imaging modalities, demonstrating its feasibility. By using a biomechanical model, we include physically-based a priori knowledge which should allow to better recover the cardiac motion from images.