Random-noise Attenuation For Seismic Data By Local Parallel Radial-Trace TFPF

Seismic exploration is an important tool in development of oil-gas and mineralresources. It is so important for the natural resource exploration that we shouldattenuate the noise and raise the signal to noise ratio (SNR) of the seismic records asmuch as possible. So seismic denoising plays an important role in increasing thequality of seismic profile data, and various denoising methods have been studied inrecent years. Unfortunately, the low SNR is a bottleneck of existing denoisingmethods in recovering the seismic events.Time-frequency peak filtering (TFPF) is widely used to suppress random noisein nonstationary deterministic band-limited signals and has been applied to seismicrandom noise reduction. In order to obtain an unbiased result, it is essential that thedesired signal has low instantaneous frequency (IF) and high degree of linearity. Inpractice, we use the Pseudo Wigner-Ville distribution (PWVD), a version ofwindowed Wigner-Ville distribution (WVD) to realize the local linearity. Thus, asuitable window length (WL) related to the frequency character is the key ofconventional TFPF to remove random noise and preserve the desired waveform.However, a single fixed WL cannot be optimal for all frequency components at thesame time. The conventional TFPF needs to be repeated several times to arrive at anoptimal window finally since we do not know the precise frequency of the signalprior to the filtering. In order to deal with the problem, this paper proposes a localparallel radial trace time-frequency peak filtering (TFPF) algorithm. Taking theapplication conditions, error resources and limitations of the conventional TFPFfully into account, we assume that the moveout of the reflected event is locallylinear and rotate the seismic reflection events to the R-T domain through the localdirections of the reflected events, to strech the apparent time and raise the linearityof the transformed seismic wavelet, which is also based on the correlation betweenthe adjacent channels. Applying the TFPF in the radial Trace transform domaincould reduce the error at the source, and problems such as amplitude degradation,width change and wave distortion could be solved. The Local Parallel radial TraceTFPF can better recover the reflected signal and raise the SNR of the seismicrecords to a higher level. Experiments on synthetic models show that the localparallel radial Trace TFPF can provide better performance in both random noiseattenuation and reflected signal preservation with a fix window length than theconventional TFPF, and the result is no longer much influenced by the windowlength as the conventional TFPF. Experiments on field data also confirmed theabove conclusion..