Abstract: The performance of synthetic aperture radar （SAR） may be affected by all kinds of radio frequency interference (RFI). Based on the RFI characteristics of high power and narrow band, a novel RFI suppression method based on eigen-decomposition is proposed for SAR. This method decomposes the eigenvectors of the received signal with RFI into interference subspace and signal subspace, and the orthogonality of the two subspaces is used to suppress the RFI signals. This paper describes the characteristics of the basic principles of eigen-decomposition, and proposes a new RFI detection method based on the MDL criterion to determine the number of main eigenvalues, and further, the eigenvectors are used to construct the corresponding interference subspace and signal subspace, and the eigen-decomposition based method for RFI suppression are proposed. The experimental results based on SAR real data shows that are also shown to verify the proposed algorithm.