A Markov random field approach for restoration of underwater acoustic images

This paper describes a method for the restoration of underwater 3D acoustic images. Beamforming is used to build a 3D map associated with another image representing the confidence (i.e. reliability) of measures. Beamforming processes backscattered echoes (received by a two-dimensional array antenna) coming from a scene previously insonified, forming two images where each pixel represents the distance from the sensor in a specified direction and the confidence (reliability) of measures. In the proposed algorithm, depth and confidence images are modelled as Markov random fields whose associated probability distributions are specified by a single energy functional. This functional embeds a-priori knowledge on the images and on the noise affecting measurements in the image formation process. Confidence measures are also considered to improve the restoration. The maximum a-posteriori (MAP) criterion is adopted to perform restoration of both images and simulated annealing is used to find the optimal MAP estimate. Experimental results show the improvement in the resulting images as compared with initial data