<p>Low-frequency signals play a crucial role in seismic inversion of thin-layer structure and reservoir prediction. However, during seismic exploration, the low-frequency signals are often contaminated, distorted, or even missing due to acquisition limitations, processing artifacts, and ambient noise. Although compressive sensing theory-based sparse inversion can partially recover low-frequency signals, the reconstruction results suffer from significant non-uniqueness. To address this challenge, we propose a sparse inversion approach incorporating spatial structural regularization to enhance low-frequency signal recovery. Due to the interference among seismic waveforms, spatial reflection structure exhibits frequency dependency. Consequently, the spatial structure estimated directly from seismic data differs significantly from the actual low-frequency spatial structure. Therefore, the proposed method estimates spatial reflection structure from seismic data in the neighboring frequency band of the low-frequency signals to be recovered, aiming to reduce the impact of frequency dependency on estimation accuracy. Subsequently, both the sparse structure of reflection coefficients and spatial structure of low-frequency signals are incorporated as regularization terms into the inversion framework, enabling geologically guided recovery of low-frequency components. The proposed method was successfully applied in the Tarim Oilfield, effectively restoring low-frequency signals and providing reliable foundational seismic data for reservoir prediction.</p>