摘要
The Cadzow technique in f-x domain was presented to attenuate random noise in a patch manner.Firstly,a Hankel matrix is constructed.Then,the Singular Value Decomposition(SVD)technique is used to decompose eigenvalues in complex field,and random noise is attenuated by replacing constant-frequency slices with the optimum weighted eigenvalues.Patch processing is utilized to solve the problem of nonstationary variation of dip angle in actual seismic data.Compared with the traditional SVD technique with the hypothesis of horizontal events,our technique does not require any assumption on the orientation of events,which does not need the relative moveout correction on the event.Meanwhile,the method is suitable for both flat and dipping events with the assumption of linear events.The model testing result indicates that the signal of the seismic data without noise will not be damaged if the number of formation dip is equal to the number of selected eigenvalues.The actual application shows that the Cadzow technique technique in f-x domain is robust for random noise attenuation.