This paper mainly considers the optimal measurement feedback control and stabilization for networked control systems (NCSs) with packet losses. The problems are involved with fundamental difficulties of separation principle and optimal estimation (conditional expectation) for multiplicative noise stochastic systems. First, the optimal estimator (conditional expectation) for NCSs with packet losses is derived and the optimal measurement feedback controller is obtained by using the maximum principle. The sufficient and necessary solvability condition of finite horizon measurement feedback control problem is first presented. Moreover, the asymptotic stability of the optimal estimator is studied. Finally, for the infinite horizon case, based on the introduction of a new Lyapunov function, which is defined with the optimal cost function, it is shown that the system is stabilizable in the mean square sense if and only if an algebraic Riccati equation admits a unique positive definite solution.