Integration of confidence information by Markov Random Fields for reconstruction of underwater 3D acoustic images

Abstract

This paper describes a technique for the integration of confidence information using a Markov Random Fields approach to improve the reconstruction process of 3D acoustical images. Beam-forming, a method widely applied in acoustic imaging, is used to arrange backscattered echoes received by a two-dimensional array antenna in order to generate two images where each pixel represents the distance (range) from the sensor plane and the related confidence of the measure, respectively. Unfortunately, this kind of images are plagued by several problems due to the nature of the signal and to the related sensing system, thus heavily affecting data quality. In the proposed algorithm, range and confidence images are modelled as Markov Random Fields and several energy formulations are devised to exploit both types of data, leading to the reconstruction and segmentation of acoustic images. Results show the better performances of the proposed method as compared with classical methods disregarding reliability information.