In this paper, a low coherence compressed channel estimation method is proposed for high mobility MIMO OFDM systems. High mobility always causes large Doppler frequency spread which costs large spectrum and time resources to obtain the accurate channel state information (CSI). As numerous recent experimental studies have shown that high mobility broadband wireless channels tend to have some inherent sparsity, compressed sensing (CS) has been introduced to utilize the inherent sparsity and reduce the CSI estimation complexity. In this paper, the coherence of CS is studied and we prove that lower coherence leads to better CS performance. An iterative algorithm is proposed to reduce the coherence by designing pilots with the known channel model before transmitted. Numerical results confirm that the proposed method has satisfied channel estimation performance in high mobility environments.