Reconstruction of textured seafloors from side-scan sonar images

The seafloor is subject to attention and exploitation in many branches of science and technology. The acquisition of seabed information using AUVs equipped with a visual system composed of cameras and sonars is becoming more common. However, although most of these sensors, such as side-scan sonar, can generate a 2-D representation of the seafloor, 3-D information, which is crucial for environmental and geological studies, is only provided by a few sensors and often only at very low resolution. A linear shape-from-shading approach is presented, which is novel in the context of side-scan sonar images. This technique aims to recover the 3-D shape of textured seabeds from side-scan sonar images. For comparison purposes, an existing propagation shape from shading approach is also considered. Results showed that both techniques yield good results in terms of similarity and accuracy, although the linear shape-from-shading technique presented superior results when noise was present in the images, a condition which is very common with side-scan sonar images.