We present a review of our work on Lagrangian Data Assimilation in a Variational Framework. The Variational Data Assimilation aims at finding an optimal trajectory, solution of a given model, that is consistant with past observations of the modelled system. The observations we are interested in assimilating are of Lagrangian type: float positions (whose dynamics contain local physical information) and images. Satellite observations of atmosphere and ocean contain a huge quantity of visual information (such as eddies, fronts) that is underused by numerical forecasting systems. The construction of the observation operator is done thanks to passive tracers, whose dynamics are interpreted as an image. The turbulent structure of this image is then compared to the observed image, both of them being represented in a set of curvlets. After a short presentation of the several steps of our method, and more especially the Image Assimilation techniques, we will present the first results on twins experiments.