Analysis of buried objects in 3D underwater acoustic images by a volumetric segmentation algorithm

To extract useful information about the buried objects contained in acoustic sub-bottom images a segmentation algorithm is mandatory. The literature on the segmentation of 3D acoustic underwater images is very limited, and, more in general, this task is still considered a challenging problem in computer vision. The volumetric segmentation method presented in this paper follows a volume growing approach, essentially a 3D extension to the traditional 2D region growing one. The volume growing operation is guided by a statistical approach based on optimal decision theory. Some pre-processing activities, e.g., filtering and enhancement, mainly aimed at preparing data to obtain good segmentation results, have also been developed.