INTRODUCTION

In multi-source acquisition, the normal spatial undersampling of CMP gathers is compounded by the fact that at each inline source location, only one of the sources is fired. This allows acquisition, which is expensive, to go more quickly, but reduces sampling at a given midpoint to every fourth offset (in the case of two sources), which may cause problems for later processing steps, such as radon multiple suppression Marfurt et al. (1996).

In this paper I subsample a multiple-contaminated 2-D marine synthetic survey to simulate a single inline from a multi-source survey. The degraded sampling hampers radon multiple attenuation. In a sequence of two steps of linear least squares, I estimate a set of local multiscale PEFs and re-estimate the removed traces. This is similar to a time-domain implementation of Spitz's algorithm 1991, with the benefit of control over the effective filter length in time Abma and Claerbout (1995). Infilling the removed data improves the separation of multiple and primary energy in radon transform space.