Seismic random noise attenuation by time-frequency peak filtering based on joint time-frequency distribution

Abstract Time-Frequency Peak Filtering (TFPF) is an effective method to eliminate pervasive random noise when seismic signals are analyzed. In conventional TFPF, the pseudo Wigner–Ville distribution (PWVD) is used for estimating instantaneous frequency (IF), but is sensitive to noise interferences that mask the borderline between signal and noise and detract the energy concentration on the IF curve. This leads to the deviation of the peaks of the pseudo Wigner–Ville distribution from the instantaneous frequency, which is the cause of undesirable lateral oscillations as well as of amplitude attenuation of the highly varying seismic signal, and ultimately of the biased seismic signal. With the purpose to overcome greatly these drawbacks and increase the signal-to-noise ratio, we propose in this paper a TFPF refinement that is based upon the joint time-frequency distribution (JTFD). The joint time-frequency distribution is obtained by the combination of the PWVD and smooth PWVD (SPWVD). First we use SPWVD to generate a broad time-frequency area of the signal. Then this area is filtered with a step function to remove some divergent time-frequency points. Finally, the joint time-frequency distribution JTFD is obtained from PWVD weighted by this filtered distribution. The objective pursued with all these operations is to reduce the effects of the interferences and enhance the energy concentration around the IF of the signal in the time-frequency domain. Experiments with synthetic and real seismic data demonstrate that TFPF based on the joint time-frequency distribution can effectively suppress strong random noise and preserve events of interest.

[1]  Yanghua Wang,et al.  Random noise attenuation using forward-backward linear prediction , 1999 .

[2]  Wenkai Lu,et al.  Local linear coherent noise attenuation based on local polynomial approximation , 2006 .

[3]  Boualem Boashash,et al.  Adaptive instantaneous frequency estimation of multicomponent FM signals using quadratic time-frequency distributions , 2002, IEEE Trans. Signal Process..

[4]  A. Scaglione,et al.  Parameter estimation of spread spectrum frequency-hopping signals using time-frequency distributions , 1997, First IEEE Signal Processing Workshop on Signal Processing Advances in Wireless Communications.

[5]  S. Cao,et al.  The second-generation wavelet transform and its application in denoising of seismic data , 2005 .

[6]  Mostefa Mesbah,et al.  Signal enhancement by time-frequency peak filtering , 2004, IEEE Transactions on Signal Processing.

[7]  Patrick Flandrin,et al.  Improving the readability of time-frequency and time-scale representations by the reassignment method , 1995, IEEE Trans. Signal Process..

[8]  G. Duncan,et al.  Median filter behaviour with seismic data1 , 1995 .

[9]  Hongbo Lin,et al.  Recovery of Seismic Events by Time-Frequency Peak Filtering , 2007, 2007 IEEE International Conference on Image Processing.

[10]  Braham Barkat,et al.  Design of higher order polynomial Wigner-Ville distributions , 1999, IEEE Trans. Signal Process..

[11]  Yanghua Wang.  Seismic trace interpolation in the f‐x‐y domain , 2002 .

[12]  Ning Wu,et al.  Noise Attenuation for 2-D Seismic Data by Radial-Trace Time-Frequency Peak Filtering , 2011, IEEE Geoscience and Remote Sensing Letters.

[13]  Mohammad Bagher Shamsollahi,et al.  Estimation of modal parameters using bilinear joint time–frequency distributions , 2007 .

[14]  Simon L. Klemperer,et al.  West-east variation in crustal thickness in northern Lhasa block, central Tibet, from deep seismic sounding data , 2005 .

[15]  Wenkai Lu,et al.  Adaptive noise attenuation of seismic images based on singular value decomposition and texture direction detection , 2006 .