Fast and Adaptive Bidimensional Empirical Mode Decomposition (FABEMD) is applied to target recognition in Synthetic Aperture Radar (SAR) image.FABEMD could fast decompose SAR images into multiple Bidimensional Intrinsic Mode Functions (BIMF)which can describe the information of targets from low to high frequencies.The index of structural similarity is employed to remove the BIMFs that are contaminated by noisesand the highly discriminative ones are kept.In the classification phase,Joint Sparse Representation (JSR) is used to represent the selected BIMFs.Experiments are conducted on the MSTAR data set to test the proposed method and the results validate its performance superiority.