The autocorrelation is calculated from a part of the trace for which we set the first and the last sample in time. The first time is here set at time zero. It is possible to choose the last sample in time. The trace's part thus selected must contain the highest possible number of multiples. The autocorrelation function of the selected trace's part is tapered by a Hanning function raised to a given power to soften the edge effects. If the processing parameters are well chosen, the autocorrelation part of the trace associated to multiples are set at zero. It is possible to modify the following parameters:

• Signal to noise ratio;
• Time of correlation window (ms);
• Prediction interval (ms).

The animation allows:

• Selecting a set of parameters of the Wiener filter so that the latter eliminates multiples on sections; a short window (300 ms) do not remove correctly multiples. An efficient window must be contain several multiples (900 ms);
• Seing that the Wiener operator is automatically adapted at each trace. The process automatically leads the prediction interval (from the initial value 99, 100, 101 and 110 ms) which is the difference in time between the maximum of the autocorrelation and the first opposite extremum value.
• Seing that the process is not sensitive to the ambiant noise.

The application of a predictive Wiener filter on the second synthetic section shows that an optimum choice of parameters allows finding the associated reflector at the water bottom and the initial reflected wave associated to the dip reflector.