This paper describes an Image Based Visual Servoing scheme for controlling a 3 DOF robot, using a system with two cameras in eye to hand configuration. Two methods of estimating image Jacobian on line by using lines as image features are proposed. The first one takes into account both the advantages of Plficker representation and properties of the stereo vision system to obtain depth sensitivity of the observed 3D lines. The second one considers direct measurements of line features in the image of an uncalibrated system. A proportional control law based on the pseudoinverse of the image Jacobian is utilized to control the task. Experimental tests have been carried out to validate the proposed methods, and extended to the analytic image Jacobian from the literature. © 2009 IEEE.