Each of these problems used a 2-D steering filter. Fomel (1999) introduced a method to construct a 3-D steering filter. Fomel formed a 3-D steering filter operator by first convolving two 2-D operators. To obtain a minimum-phase filter he performed spectral factorization for each dip component (px , py) pair in the data, significantly increasing the cost of constructing the operator. In order for the resulting filter to spread information over significant distances, a large number of filter coefficients must be used, increasing the cost of each iteration.
In this paper I present an alternative construction method. I show how a 3-D steering filter can be produced by cascading two 2-D steering filters. The cascaded approach does not provide as accurate a dip discrimination as that in Fomel's approach but it does not require the expensive spectral factorization, and the resulting filters are much smaller and less expensive to apply. I show how this method can accurately characterize a large range of dips and that it is accurate enough for a wide class of applications. In addition, I apply it to a simple synthetic missing data problem with very encouraging results.