This paper considers the problem of position control of robot manipulators via visual servoing in presence of uncertainties associated with the robot dynamics and the vision system. Specifically, an adaptive algorithm is designed to compensate for the uncertainties associated with the mechanical model of the robot manipulator and the intrinsic parameters of the camera. We propose a control law which requires for implementation only joint and image coordinates, also a velocity observer is designed. This is achieved by combining a previous control algorithm designed originally to work Cartesian coordinates and an adaptive control law, to a version which uses image coordinates instead. Simulation results show the good performance of the complete system.