The downward-looking three-dimensional synthetic aperture radar (DL 3D SAR) imaging method based on compressed sensing could complete the high-resolution imaging by using the sampling data below the Nyquist sampling rate. However, the existing DL 3D SAR imaging methods which used for cross-track profile reconstruction are based on the single measurement vectors (SMV) model and has defects of time-consuming and noise-affected. Based on the multiple measurement vectors (MMV) model, the order of cross-track and along-track in the DL 3D SAR imaging process are exchanged. By using the same sparse structure in cross-track domain for different along-track measurements, the DL 3D SAR imaging method based on MMV model is proposed. The proposed method is superior to the SMV-based method both on time-consuming, reconstruction accuracy and anti-noise performance, and the effectiveness is verified by the simulation experiments.