The complex bidimensional empirical mode decomposition (C-BEMD) was applied to target recognition of synthetic aperture radar (SAR) image. As an extension of traditional BEMD to complex domain, C-BEMD could directly process the complex SAR images (including the amplitude and phase information). C-BEMD was employed to decompose SAR images to obtain multi-layer bidimensional intrinsic mode functions (BIMF), which could reflect the time-frequency properties of images. These BIMFs had individual description capabilities, which reflected the target characteristics from different aspects. Also, they shared inner correlations, which were originated from the same target. The classification algorithm was developed based on the joint sparse representation, which used the inner correlations to improve the representation precision. The multi-class SAR images in the MSTAR dataset were used to test and validate the proposed method. The results confirm its reliable recognition performance under the standard operating condition (SOC) and extended operating conditions (EOC).