This paper concerns the stability analysis of a new visual servoing approach which is invariant on camera intrinsic parameters. Contrarily to standard methods, the invariant visual servoing approach can be used with a zooming camera or when the reference image is learned with a camera different from that used for servoing. Even if the error computed in an invariant space does not depend on the camera intrinsic parameters, they are needed to estimate the interaction matrix which links the camera velocity to the displacements of the features in the invariant space. Thus, calibration errors can affect the stability of the control law. For this reason, it is important to study the robustness of the proposed vision-based control with respect to uncertainties on the parameters of the system.