Multiples Suppression Using Non Linear Filtering In (τ, P) Domain

In seismic processing, the suppression of multiples is often considered as an important step. Initially the conventional filter for suppression of multiples in the (f, k) domain has proved efficient namely for the far offsets. However, due to some arising geological difficulties (especially when the multiple energy is superposes to that of the primary reflection) this filter is no more appropriate. New techniques are then required. Non linear filtering allows us to define automatically the zone to eliminate, even energies of events (multiple and primary reflections) are superposed or are overlapped. To release this filter, the Butterworth gain function was used to compare the energies of the primaries and multiples reflections. The non linear filter is then applied using the original model and the predicted multiple model. This latter leads to better results. This was the subject of a first work. Filtering by the linear τ-p transformation has also given good results. However, the parabolic transformation gives much better results. Its main advantage is that a reflection in the (x, t) domain, transforms into a point in the (τ, p) domain making its elimination much easier to perform when the reflection is a multiple. In this work, we deal with the non linear filter using the predicted multiple model in (τ, p) domain (linear and parabolic).