Abstract-In traditional over-the-horizon radar (OTHR), multipath propagation due to the multilayer ionospheric structure always deteriorates the detection performance. The properties of multipleinput multiple-output (MIMO) radar technique, which transmits wide beams with low gain at the transmitter and achieves receiver beam-forming to obtain narrow beams with high gain, make it an ideal choice for OTHR to detect target through multi-layer ionosphere and suppress strong clutter. This paper investigates the assumption of a two-layer ionospheric model and proposes a two-step MaxMin algorithm based on the mutual information theory to optimize MIMO-OTHR waveform so as to suppress clutter, interference and noise. The first step is to maximize the mutual information between the echo and target response from the same direction of arrival (DOA) in order to reduce the impact of noise. The second step is to minimize the mutual information between the echoes from different DOAs, in order to suppress the clutter and interference by reducing the correlation of the echoes from the different DOAs. Numerical experiments validate that this algorithm can improve range resolution and detection probability significantly. Experiment results also demonstrate that the previously harmful multipath propagation can be utilized to enhance the detection performance in MIMO-OTHR.