Automated breast ultrasound imaging systems provide the ability to combine the high anatomical detail of preoperative imaging with real-time intraoperative imaging. Ultrasound imaging involves interaction between the transducer and the patient, which requires a compliant behaviour of the robot in manipulating the transducer. We present an energy-based control strategy augmented with a visual servoing algorithm to improve the image quality online. An impedance controller is extended with energy- and power-based safety metrics to adjust the dynamic behavior of the robot and with energy tanks to ensure passivity. The quality of the ultrasound images is represented by a confidence map that is computed online, following a model of ultrasound propagation in soft tissue. Experiments were conducted with the 7 degree-of-freedom KUKA LBRMed articulated robot. First, a port-Hamiltonian model of the robotwas created to design and test the controller in simulation. Second, the controllerwas transferred to the robot controller for experimental validation of the proposed control strategy. Simulational and experimental results have validated the controller design in interaction with the environment. Experiments with an ultrasound phantom have demonstrated that the visual servoing algorithm corrects the probe pose in case of (partial) contact loss between the transducer and the skin surface. This significantly increases the global image quality of the 3D image of the breast. Our approach has resulted in a very promising control strategy for autonomous ultrasound scanning that can be integrated in an automated breast biopsy procedure.