Mobile manipulation in service robotic applications requires the alignment of the end-effector with recognized objects of unknown pose. Image based visual servoing provides a means of model-free manipulation of objects solely relying on 2D image information. This contribution proposes a visual servoing scheme that utilizes the pixel coordinates, scale and orientation of the Scale Invariant Feature Transform (SIFT). The control is based on dynamic weighted moments aggregated from a mutable set of redundant SIFT feature correspondences between the current and the reference view. The key idea of our approach is to weight the point features in a way that eliminates or at least minimizes the undesired couplings to establish a oneto-one correspondence between feature and camera motion. The scheme achieves a complete decoupling in 4DOF, additionally the visual features in 6DOF are largely decoupled except for some minor residual coupling between the horizontal translation and the two rotational degrees of freedom. The decoupling results in a visual control in which the convergence in image and task space along a particular degree of motion is not effected by the remaining feature errors. Several simulations in virtual reality and experiments on a robotic arm equipped with a monocular eye-in-hand camera demonstrate that the approach is robust, efficient and reliable for 4DOF as well as 6DOF positioning.