Image processing techniques provide access to full field measurements of different thermomechanical data (strain; strain-rate, Wattrisse et al., J Exp Mech, 41:29-38, 2001; temperature, Chrysochoos and Louche, Int J Eng Sci 38:1759-1788, 2000. These techniques have become increasingly useful for obtaining fine and local descriptions of material properties. As they can measure complete thermal and mechanical fields, they can be used to identify several parameters of constitutive equations during a single deformation process using specifically designed heterogeneous tests (Grédiac, Composites: Part A 35:751-761, 2004). In Geymonat and Pagano (Meccanica 38:535-545, 2003), surface strain fields obtained by digital image correlation were used to identify the distribution of elastic parameters and stress fields by minimizing a given energy functional. In this paper, the previous method is improved through a relevant choice for stress approximation, and then extended to a wider class of elastoplastic materials. Its reliability is then checked through applications on simulated data obtained under small perturbation and plane stress assumptions. In particular, the robustness of the method with respect to measurement noise is studied on the basis of numerical data. An experimental application to heterogeneous material identification is finally proposed.