Adaptive Subtraction Based on Joint Matching Filtering and Its Application to Seismic Multiple Attenuation

Conventional adaptive subtraction methods for multiple attenuation rely solely on predicted multiples, resulting in poor removal at near offsets. To address the issue, we develop an adaptive subtraction based on joint matching filtering (ASMF) algorithm. It is developed under the premise that the common midpoint (CMP) gathers consist of primaries, multiples and random noise. The algorithm assumes that accurate primaries can be obtained from the far-offset traces of the CMP gathers. Another assumption is that the near-offset and far-offset data in the gathers differ mainly in terms of fluids, amplitude variation, frequency attenuation, and signal-to-noise ratio (SNR). The method further assumes that corresponding primary events remain approximately aligned after normal moveout (NMO) correction and that angle-dependent reflectivity variations are limited and do not involve polarity reversals. After conventional separation, selected far-offset traces of the preliminary primary estimate are stacked to estimate a zero-phase wavelet for reflection coefficient inversion. The resulting reflection coefficient is convolved with trace-dependent wavelets to obtain the predicted primaries. Based on the predicted primaries and multiples, the matching filtering operators for both components are jointly calculated under the framework of morphological component analysis (MCA). Incorporating the predicted primaries into joint fitting can help improve the separation of primaries and multiples. Numerical experiments and field data tests demonstrate that, compared with the Radon transform and conventional adaptive subtraction, the proposed algorithm can effectively improve the accuracy of multiple attenuation.