The compressed sensing (CS) based imaging methods for tomography SAR perform well in the case of large number of baselines. Unfortunately, for the current tomography SAR, the baselines are obtained from many multi-pass acquisitions on the same scene, which is expensive and can be severely affected by temporal decorrelation. In order to reduce the number of baselines, a novel strategy for tomography SAR imaging by introducing the block-sparsity theory into the imaging processing is proposed in this paper. Using neighboring pixels information in reconstruction, the proposed method can overcome the imaging quality limitation imposed by the low number of baselines. The results with simulation data under the additive gaussian noise case are presented to verify the effectiveness of the proposed method.