In this paper, we present a novel method for assimilating geometric information from observed images. Image assimilation 6 technology fully utilizes structural information from the dynamics of the images to retrieve the state of a system, and thus to better predict its evolution. Additionally, the level set method, which describes the evolution of the geometry shapes of a given system, is taken into account to include the dynamics of the images. This method differs from previous methods of image assimilation in that it takes advantage of Lagrangian information in an Eulerian numerical framework. In our numerical experiments, we apply this technique of image assimilation based on the level set method to an oil pollution problem, to calibrate the initial contours of oil pollutants and to identify diffusion coefficients of the model. Topological merging and breaking of oil slicks are well defined and easily performed by this proposed approach. The results show good agreement between simulated values and observed images.