The reliability and robustness of image-based visual servoing systems is still unsolved by the moment. In order to address this issue, a redundant and cooperative 2D visual servoing system based on the information provided by two cameras in eye-in-hand/eye-to-hand configurations is proposed. Its control law has been defined to assure that the whole system is stable if each subsystem is stable and to allow avoiding typical problems of image-based visual servoing systems like task singularities, features extraction errors, disappearance of image features, local minima, etc. Experimental results with an industrial robot manipulator based on Schunk modular motors to demonstrate the stability, performance and robustness of the proposed system are presented.