Layover is a difficult problem in SAR imaging technology. Airborne multi-baseline SAR systems are constrained by non-uniform baselines that are also few in number, which means the measured scattering point height is characterized by unacceptable error and poor resolution when using traditional spectrum estimation methods. To solve these problems, in this paper, we combine fourth-order cumulant statistical properties with traditional spectrum estimation methods. With its blind Gaussian noise characteristic, non-uniform virtual array extension of the fourth-order cumulant, and traditional Capon and MUSIC spectrum estimation approaches, the proposed method can effectively remove the Gaussian noise, while also improving the measurement accuracy and resolution of the scattering point heights. The simulation and real-data experimental results validate the effectiveness of our proposal.