This paper presents a new image-based visual servoing approach to control a robotic system equipped with an ultrasound imaging device. The presented method allows an automatic positioning of the probe with respect to an object of interest. Moments-based image features are computed from three orthogonal ultrasound images to servo in-plane and out-of-plane motions of the system. An efficient segmentation method, based on graph cut strategy, is proposed to extract the object contour in each image plane. Simulation results demonstrate that this approach improves upon techniques based on a single 2D US image in terms of probe positioning. Our method was also validated from robotic experiments performed on an ultrasound phantom with the use of a motorized 3D probe that provides the three US images.