Coherent noise attenuation for passive seismic data based on iterative two-dimensional model shrinkage

Passive seismic source imaging can be utilized to recover geophysical information from subsurface ambient noise. Compared with conventional active seismic exploration, passive seismic source imaging is cost-effective and environmentally friendly. However, passive data acquisition cannot easily satisfy the theoretical condition, leading to noised virtual-shot gathers. Furthermore, coherent noise limits the application of passive source data. Although image quality improvement techniques for passive source data have recently attracted considerable interest, the denoising problem for virtual-shot gathers is seldom considered. In this study, we propose an iterative denoising approach for passive seismic data. The criterion used to extract useful signals is the difference between the wavefield similarity of useful events and the coherent noise in various gathers, i.e., the common shot gather and common receiver gather. We adopted local similarity to measure the similarity level and extract major useful events. However, the close local similarity between weak events and coherent noise may cause signal leakages and singular noise residuals. We incorporated an iterative two-dimensional model shrinkage algorithm into the denoising process to suppress the singular noise residual and highlight useful events. The proposed approach can overcome the limits of strong coherent noise in virtual-shot gathers, which can extend the choice range for data processing. Synthetic and field examples demonstrate a promising coherent noise attenuation performance, illustrating the effectiveness and feasibility of the proposed method. The denoised migrated section exhibits a smaller depth error and higher quality.