This paper introduces a methodology to reconstruct underwater environment using two images of the same scene, acquired by an acoustical camera from different points of view. The final target of the work is to produce a full 3D representation of the observed environment to improve its exploration and analysis. Indeed, as the DIDSON acoustic camera provides sequences of 2D images (distance and azimuth), the challenge consists in determining the missing elevation information about the observed scene in order to reconstruct (x, y, z) models, through the computation of the geometrical transformation between the acquisition view points, using image information only. Our research work is divided in two important steps. The first step which is feature point extraction allows robust and shape representative point extraction. The second step presented in this paper uses these specific points appearing on two images and paired accordingly, to determine camera motion (rotation and translation) between the two acquisitions, and points missing elevation in order to reconstruct the observed scene. Due to the problem high-dimensional search space (6 camera motion parameters plus one elevation per pair of points), we propose to achieve the search using CMA-ES optimization algorithm. This stereovision-like optimization procedure assumes a known camera model. The first topic in this paper tries to check the good behavior of the supposed camera model in order to be sure that extracted points from images are robust enough and not affected by extra camera distortions. A set of DIDSON images have been acquired in the Laval University pool and used to perform such a verification, with various objects (wooden boxes and grid) observed from different points of view. Finally, using extracted pairs of points coming from two images, the proposed algorithm is able to retrieve the local relative geometry of the observed scene through the estimation of the missing elevations.