Multiple attenuation via Blind Source Separation methods

Multiple elimination is a very important step in seismic data processing. It is commonly undertaken by methods using least squares (LS) filters to perform adaptive subtraction. However, these techniques usually cause some collateral damage on primaries that intersect with multiples. This is due to the LS filters whose optimization criterion relies on removing as much energy as possible around the predicted multiples. In recent works, Blind Source Separation (BSS) methods have been applied to geophysics and have shown interesting results for multiple extraction. These techniques are able to identify and separate primaries from multiples without any adaptive subtraction, hence minimizing error when primaries and multiples overlap. In view of these initial results, we present in this paper a study on the application of BSS techniques to the problem of multiple extraction. Our study encompasses the analysis of different BSS methods and their application to a number of scenarios, considering both synthetic and real data.