In existing methods of ISAR imaging with MMV (Multiple Measurement Vector) SBL (Sparse Bayesian Learning), the HRRP is always interpolated radially to realize the column-to-column correspondence between HRRP matrix and scattering coefficient matrix, then scattering coefficients are estimated with MMV SBL algorithm. In order to enhance the ability to suppress noise and eliminate stripe interference, a novel MMV SBL algorithm based on bidirectional interpolation (BI-MSBL) is proposed for ISAR imaging in this paper. Besides the radial interpolation and then estimation of scattering coefficients conventionally, the HRRP is also interpolated transversally to realize the row-to-row correspondence between HRRP matrix and scattering coefficient matrix, then scattering coefficients are also estimated with MMV SBL algorithm. So there are two ISAR images can be gained respectively from radially interpolated HRRP and transversally interpolated HRRP. Finally, the two obtained ISAR images are fused to get the final ISAR image. In the experiments, simulated HRRP data and measured HRRP data under different SNRs are used to test the performance of different algorithms, including the conventional R-D (Range-Doppler) algorithm, the MMV SBL algorithm, the PC-MSBL algorithm, and the BI-MSBL algorithm proposed in this paper. Through the comparison among the experimental results, it can be found that the BI-MSBL algorithm is more effective than the other three algorithms in suppressing noise, eliminating stripe interference, and enhancing the clarity of ISAR images. As a shortcoming, the BI-MSBL algorithm needs to work out two estimations of the scattering coefficients, so it needs twice the computation time as that of the MMV SBL algorithm.

在现有的MMV（Multiple Measurement Vector）SBL（稀疏贝叶斯学习）ISAR成像方法中，总是径向插值HRRP以实现HRRP矩阵与散射系数矩阵的列到列对应关系，然后用MMV估计散射系数SBL 算法。为了增强抑制噪声和消除条纹干扰的能力，本文提出了一种基于双向插值的新型MMV SBL算法(BI-MSBL)用于ISAR成像。除了常规的径向插值然后估计散射系数外，还对HRRP进行横向插值，实现HRRP矩阵与散射系数矩阵的行间对应关系，然后用MMV SBL算法估计散射系数。因此可以分别从径向内插HRRP和横向内插HRRP得到两幅ISAR图像。最后将得到的两幅 ISAR 图像进行融合，得到最终的 ISAR 图像。实验中，利用不同信噪比下的模拟HRRP数据和实测HRRP数据来测试不同算法的性能，包括常规RD（Range-Doppler）算法、MMV SBL算法、PC-MSBL算法和BI-本文提出的MSBL算法。通过实验结果的对比，可以发现BI-MSBL算法在抑制噪声、消除条纹干扰、增强ISAR图像清晰度等方面比其他三种算法更有效。作为一个缺点，BI-MSBL 算法需要计算散射系数的两个估计，