Discussion

In complex areas, the image space is an attractive domain to attenuate the multiples, provided that we can design a Radon transform to separate them from the primaries. The presence of large salt bodies and in particular their steeply dipping salt flanks in the cross-line direction severely distort the imaging of the multiples. A full 3D migration is necessary to image the multiples. In order to discriminate between primaries and multiples, and to attenuate the multiples, we need to compute prestack image gathers as a function of subsurface offset or aperture and azimuth angles. Computing prestack images is expensive, but they can be used not only to attenuate the multiples but also to assess the accuracy of the migration velocity model and even to back project residual moveout information into velocity corrections. The expense of computing the full prestack image (inline and cross-line subsurface offsets and from them angle gathers as a function of aperture angle and reflection azimuth) is thus well worth.

At a relatively small additional cost (an additional convolution), it is attractive to use the image space version of SRME to get an initial estimate of the multiple model Artman et al. (2007). I expect, however, that diffracted multiples, and specular multiples from reflections with a significant cross-line dip components, will not be accurately modeled. For these multiples, exploiting their particular behavior in terms of their residual moveouts as a function of aperture angle and reflection azimuth, is the best way to attenuate them.