In this paper, we present an ultrasound (US) visual servoing to control a robotic system equipped with a US probe. To avoid the difficult and time-consuming image segmentation process, we develop a new approach taking as visual input directly the intensity of the image pixels. The analytic form of the interaction matrix that relates the variation of the intensity features to the motion of the probe is established and used to control the six degrees of freedom (dof) of the robotic system. Our approach is applied with a 2-D and a 3-D US probe, and the results that are obtained with both sensors are compared in simulation. The 2-D probe shows good performances for tracking tasks and the 3-D one, which ensures a larger domain of convergence, is more particularly used for positioning tasks. The intensity-based approach is validated through experimental results performed with a realistic abdominal phantom and with animal soft tissue.