Optimizing the measurement matrix can improve reconstruction performance in compressed sensing. In this study, we study the measurement matrix optimization method regarding its application to the Two Dictionaries Orthogonal Matching Pursuit (TDOMP) algorithm. The TDOMP is a modified OMP, which uses a matching matrix with low cross-coherence to identify the correct atoms of the sensing matrix. The proposed optimization method is based on alternative projection technique to construct the measurement and matching matrices with low cross-coherence to improve the performance of the TDOMP. Experimental results verify the effectiveness of the proposed method.