When the sensing matrix holds the restricted isometry property (RIP), the compressed sensing-based OFDM channel estimation can greatly reduce pilot overhead and approximately ideally recover the channel state information (CSI) as well by exploiting the sparsity of channel impulse response (CIR). In this paper, an optimal deterministic pilot pattern is proposed under the principle of minimizing the mutual correlation coefficient of the sensing matrix. With the aid of the designed optimal definite pilot pattern, the mutual correlation coefficient of the sensing matrix is analyzed, and the mean square error (MSE) performance of orthogonal matching pursuit (OMP) algorithm is simulated for OFDM systems with large amount of virtual subcarriers. Both the analysis and the simulation results show that the designed optimal definite pilot pattern outperforms the other reference pilot patterns.