Aimed at the problem that the high computational complexity of airborne multipleinput multipleoutput(MIMO) radar spacetime adaptive processing (STAP) technique exists in using the sparse method to recover the clutter spectrum, a dictionary dimension reduction method is proposed. In this method, the training samples are directly used to estimate lowresolution clutter spacetime spectrum. On this basis, the resolution of spectrum is improved by using the iterative formula of FOCUSS algorithm. Then the Capon spacetime power spectrum values of each atom are calculated and the atoms with large spectral values are selected to form the reduceddimension dictionary. When there is doppler ambiguity, the prior knowledge is used to exclude the atoms selected because of doppler ambiguity, and the remaining atoms met the requirements are taken out to form the reduceddimension dictionary. The simulation results show that the reduceddimension dictionary can completely cover the clutter ridge, and simultaneously can effectively improve the operation efficiency while maintaining the SINR output performance of the algorithm.