Conclusion

We applied a new t-x domain, pattern-based signal/noise separation technique to a 2-D line contaminated with multiples. This technique differs from the Wiener-like method because the data components are not filtered but rather predicted. The goal of this work was to show (1) that the noise attenuation can be formulated in a totally different way using a new fitting goal, (2) that this new formulation leads to a proper subtraction of the noise components, and (3) that so far, this method is comparable in efficiency to the Wiener-like method. The first results are encouraging. Yet, we have not explored the possibilities offered by the regularization to improve the signal-noise separation.