Unavoidable random noise not only lowers the signal-to-noise ratio (SNR) of seismic data but also decreases the accuracy of dynamic and static corrections, which leads to degradation of the final data quality. Furthermore, it is often a problem in geophysical data visualization because it obscures fine details and complicates identification of profile features. It is very significant to reduce random noise and protect useful information. Aiming at the poststack random noise elimination, the paper introduces an adaptive wiener filter. On the basis of comparing theoretical model results, the method can perfectly filter background noise with near half energy of useful information, improving the whole visage of seismic data. Application of the method to poststack data from the Songliao basin in China demonstrates the effectiveness of the method.