Abstract: The modern underwater acoustic equipment, like multibeam sonar, is able to provide the data which allows for unambiguous localization and also 3D visualisation of targets of different kinds, including wrecks, mines and other submerged objects. The large amount of information that seems to be useful for this purpose, could also be extracted from side-scan sonar data, which has been collected over years within different survey regions. In this context, the paper concerns the development of side-scan data processing method for 3D reconstruction of seafloor and its imaging. In this paper, two novel approaches are proposed. The first utilises the more accurate models of backscattering, including the differences in signatures of several types of objects. The second is based on definition and investigation of some quantitative descriptors of geometric features expected to be found in images of localised objects of artificial origin. The preliminary results obtained by application of these techniques to side-scan sonar data processing will be discussed in this paper.