Abstract: Singular value decomposition (SVD) is found to be extensive application in attenuating random noise of seismic data processing. Focusing on the random noise attenuation, seismic data matrix is reselected by event shapes determined with polynomial expression, the data matrix reselected is then filtered by SVD in order to attenuate random noise. The study of SVD reveals the principles of SVD and its filtering,which offer theoretical basis for further application of SVD in geophysical prospecting. Application with real data shows that this method used for removing noise is of flexibility, high reality and high resolution. Also, it has significant effects on sudden pulse interference, offside reflection and other reverse interferences.

Key words: singular value, random noise, polynomial fit, signal/noise ratio