Adaptive subtraction of multiples

Chapter illustrates on a multiple attenuation problem how the Huber norm can help to better separate primaries from multiples. Adaptive subtraction of multiples is a two steps process where multiples are first predicted, leading to an accurate model of the noise, and then adaptively subtracted from the data by estimating matching filters. When weak multiples are present in the surroundings of strong primaries, multiples can be matched to primaries when the $\ell^2$ norm is utilized for the filter estimation step. In this Chapter, I prove that the Huber norm provides a robust measure for computing filter coefficients that is less sensitive to the relative strengths of both multiples and primaries, thus preserving the signal better than the $\ell^2$ norm.