Many current and future applications of underwater robotics require real-time sensing and interpretation of the environment. As the vast majority of robots are equipped with cameras, computer vision is playing an increasingly important role in this field. This paper presents the implementation and experimental results of underwater StereoFusion, an algorithm for real-time 3D dense reconstruction and camera tracking. Unlike KinectFusion on which it is based, StereoFusion relies on a stereo camera as its main sensor. The algorithm uses the depth map obtained from the stereo camera to incrementally build a volumetric 3D model of the environment, while simultaneously using the model for camera tracking. It has been successfully tested both in a lake and in the ocean, using two different state-of-the-art underwater Remotely Operated Vehicles (ROVs). Ongoing work focuses on applying the same algorithm to acoustic sensors, and on the implementation of a vision based monocular system with the same capabilities.