Multisensor fusion

Applications involving remotely-sensed images arouse the interest of a large user community. The paper focuses on the fusion of multisensor images available from satellite onboard sensors and in particular on image processing. The physico-diversity of image dimensions allows combination as an aid to extracting relevant information.. Such dimensions as spatial, spectral, temporal, angular or polarimetric can be cleverly merged. Data assimilation models corresponding to either a confrontation model or an incorporation model are described within three architectures: centralised, de-centralised and mixed. The quality of the result depends on the assessment of several sources of error. Synergism as a function of data interaction and errors management is discussed within these architectures. Certain image dimensions requisite to improve data fusion techniques are nowadays taken into account as from the conception of a sensor programme. In conclusion, the fusion of data as an aid to define new value-added products is presented, which allows us to consider this technique as being one of the future trends in remote sensing.