Automatic Rectification of Side-scan Sonar Images

The authors present a novel procedure for the automated rectification of side-scan sonar images. The traditional assumption of a flat seabed for the computation of the ground projection from the slant-range sonar data can later result in difficult geo-referencing and registration problems. Accurate rectification using actual seabed topography is required for precise integration of side-scan images into a high-resolution mosaic, which depends on the availability of seabed altitude maps for the area of interest. As an alternative the authors propose a procedure for the estimation of the most probable seabed topography given the acquired side-scan image and a simple scattering model. An Expectation-Maximization framework is constructed to iteratively determine the model parameters that best approximate the observed image.