Filtering > Wiener filter: predictive deconvolution

Seismic sections can contain multiples. Multiples are repetitions in time of primary seismic signals. The amplitude of multiples decreases with time. In order to filter multiples, we can use deconvolution operators called predictive deconvolution which require an estimation of the repetition time of multiples. Any deconvolution is sensitive to noise. A method classicaly used and quite robust to noise is the predictive Wiener filter. Wiener filter allows passing from a given signal to a desired signal. The Wiener filter is the filter which minimizes (with the least squares evaluation) the difference between the desired signal and the signal estimated by the filter. In order to estimate multiples by Wiener, the desired signal is the input signal delayed by the repetition time of multiples called prediction interval. The video shows the use of the predictive Wiener filter to eliminate multiples.