In this paper, we address the problem of characterizing objects in 3D sonar images obtained by a multibeam echo-sounder. Compared to classic 2D images from monobeam echo-sounder, new descriptors must be found for 3D images, which give more detailed information on objects. Viewing objects patterns as realizations of spatial point processes, descriptive statistics providing a joint characterization of visual content and spatial organization of the image are investigated. This method is then applied to classify fish schools on 2D and 3D sonar images. Reported experiments illustrate the relevance of the proposed descriptors.