Abstract: Multiple attenuation is a research focus in marine seismic data processing. Due to the advantages of high accuracy and low model dependence, the surface-related multiple elimination method achieves multiple attenuation through multiple prediction and adaptive subtraction, so it is widely used in the industry. However, the conventional adaptive subtraction method is not applicable at the primary and multiple intersections, and the phenomenon of primary damage and multiple residuals will increase the difficulty of subsequent seismic processing and interpretation. Based on previous research, in this paper, an adaptive multiple subtraction method based on empirical low-rank representation is proposed. We use the empirical mode decomposition method to improve the conventional low-rank representation method. By adaptively decomposing the seismic signal into low-rank subsets with high signal-to-noise ratio, simple inclination, and smooth event phase, the parameter selection of the local window during the low-rank representation processing is optimized, that is, to reduce the complexity of the dip component of the seismic signal; then the conventional adaptive subtraction method in different low-rank subsets is adopted to avoid the event intersections in the adaptive subtraction, and each subset is reconstructed to improve the multiple attenuation effect under the premise of ensuring the calculation efficiency. In order to verify the effectiveness of the proposed method, we apply it to the synthetic and field examples, the results are better than those of the conventional methods, and the noise sensitivity example proves the robustness of the proposed method.

Key words: low-rank representation, empirical mode decomposition, multiple matching, multiple subtraction