Surface-related multiple attenuation (SRMA) can effectively remove multiples from seismic data when other (e.g., Radon transform) methods have difficulty. SRMA generally includes two steps: multiple prediction (or multiple modeling) and adaptive multiple subtraction (AMS). In some cases, adaptive subtraction is the main challenge for the success of SRMA. Adaptive subtraction is often posed as a least-squares minimization problem that minimizes the energy difference between the original input traces and the modeled multiple traces. The minimum-output-energy approach can be implemented by either single- or multichannel matching filters. These methods assume orthogonality between the primaries and multiples, meaning that the two classes of events are uncorrelated.