Abstract : In the last decades satellites datas quantities and qualities have increase. Currently, satellite images sequences, which could provide a lot of information on atmospheric or oceanic system dynamic, are underused in numerical weather forcasting. The aim of this work is to provide new methods to use this information. In this study, we address the topic of monitoring the initial state of a system through the variational assimilation of indirect images sequence. The feasability of the approach is demonstrated on twin experiments realised with a shallow-water model. The influence of the distance definition between two images sequences is studied in presence of data noise.