In this paper, we consider a leader-following consensus problem for networks of continuous-time integrator agents with a time-varying leader under measurement noises. We propose a neighbor-based state-estimation protocol for every agent to track the leader, and time-varying consensus gains are introduced to attenuate the noises. By combining the tools of stochastic analysis and algebraic graph theory, we study mean square convergence of this multi-agent system under directed fixed as well as switching interconnection topologies. Sufficient conditions are given for mean square consensus in both cases. Finally, a numerical example is given to illustrate our theoretical results.