Deep seafloor characterization with multibeam echosounders using image segmentation and angular acoustic variations

Because the use of multibeam echo-sounder imagery in seafloor identification is constantly increasing, a semi-automatic mosaic interpreter is presented. It is based on the statistical and acoustical properties of the image pixels, and relies on the use of Markov random fields image models within a Bayesian framework for partitioning mosaics into homogeneous regions. Further, the authors introduce a Gibbs distribution model of the original image for computing its maximum a posteriori estimate. Effects of backscattering angular variations are compensated by injecting a first estimate of these into the calculation. Segmentation result of low-frequency multibeam mosaic is presented and compared to geological interpretation