In airborne bistatic multiple-input multiple-output radar systems, the transmit angle domain can be exploited to improve the performance of clutter cancellation. However, the dimensions of the clutter covariance matrix (CCM) are increased and more independent and identically distributed (iid) training samples are required to estimate the CCM. In this paper, a sparse recovery based dimensional reduced method is proposed to cancel the clutter without training samples. Unlike exist direct data domain sparse recovery (D3SR) method, the proposed method only uses partial grids in space-time domain to reconstruct the CCM and the computation load is then reduced. Simulation and analysis confirm the effectiveness of the proposed method.