Diffractions with good illumination carry valuable subwavelength information below the Rayleigh limit and can be used for high-resolution imaging of small-scale discontinuous features. As diffractions are characterized by weak energy, diffracted wavefield separation is a key step before imaging of discontinuities. The traditional multichannel singular-spectrum analysis (MSSA) for diffraction separation considers the constant rank strategy. It is challenging to estimate the appropriate rank for various data. In this study, we propose a robust adaptive rank-reduction method for 3D diffraction separation and imaging. It utilizes the low-rank characteristics of reflections in the common-offset or poststack domain based on the MSSA algorithm. The method considers the kinematic and dynamic differences of reflections and diffractions, and combines the energy intensity and smoothed curvature to accurately estimate the rank thresholds as well as the ranks of the reflections and diffractions. Three-dimensional multilayer and complex synthetic examples are used to demonstrate the feasibility of the proposed method in removing reflections and separating diffractions and achieving high-resolution imaging with diffractions, which is helpful for seismic data interpretation.

具有良好照明的衍射携带低于瑞利极限的有价值的亚波长信息，可用于小尺度不连续特征的高分辨率成像。由于衍射的特点是能量较弱，因此衍射波场分离是不连续成像之前的关键步骤。用于衍射分离的传统多通道奇异谱分析 (MSSA) 考虑了恒定秩策略。估计各种数据的适当等级具有挑战性。在这项研究中，我们为 3D 衍射分离和成像提出了一种鲁棒的自适应降阶方法。它利用基于 MSSA 算法的共偏移量或叠后域中反射的低秩特征。该方法考虑了反射和衍射的运动学和动力学差异，并结合能量强度和平滑曲率来准确估计等级阈值以及反射和衍射的等级。通过三维多层复杂的合成实例，证明了该方法在去除反射和分离衍射以及实现高分辨率衍射成像方面的可行性，有助于地震资料解释。