Abstract:Noise elimination before and after stacking is necessary in processing low S/N seismic data. Signal-noise separation technique based on linear theory fails to meet the requirements of seismic data processing when seismic noises are powerful. Signal-noise separation using mid-value correlative filtering has been developed by analysing some existing noise elimination techniques. This method separates signal from noise by taking two essential steps:1)to obtain a mid-value sequence in the scope of apparent dip angle,and 2) to derive the optimum mid value from the midvalue sequence by correlation analysis,the optimum mid value being taken as predictive coherent signal. The method is convenient and applicable to signal-noise separation in severe noise background;besides,it brings high fidelity of signal and so is a desirable method. for processing low S/N seismic data in complicated area.