In view that the transition probability matrix (TPM) can only get part of the measurement data, the H control problem of discretetime Markov jump systems is considered by using an optimal estimation method. Firstly, an optimal estimation of TPM is obtained. Then, several sufficient existence conditions for modedependent H controllers are established in terms of linear matrix inequalities (LMIs). Comparing with the traditional results developed by the robust method, the proposed results in this paper are less conservative. Finally, the utility and superiority of the proposed method are verified by numerical examples.