Research On High Frequency Random Noise Suppressing Methods Of Seismic Data

In view of the contraction between image resolution and signal to noise ration in high resolution seismic imaging method,based on time-frequency analysis,and using prestack seismic signal with time-varying spatial-variant properties,this paper conducted systematic researches,respectively from the wavelet transform,generalized S transform,empirical mode decomposition,on the basic principles,algorithms and procedures of prestack high frequency noise suppression.The results show that,using the difference between effective signal and the high frequency noise in wavelet coefficient energy,the two-dimensional discrete wavelet determin the different decomposition subband threshold size according to data driving method,to meet the characteristics in energy,space variable distribution among the non-stationary seismic data.Through using continuous hard threshold function,this method can,smoothly and continuously,remove low energy wavelet coefficients,so as to suppressing high frequency noise.According to the different distribution of efficient signal and the high frequency noise with different seismic propagation time,generalized S transform reasonably determined the frequency pressing range of seismic data in different time,in order to eliminate the high frequency noise in prestack data.This method,with time varying and space varying filtering properties,can be fully retained the effective signal energy,so it is very suitable for prestack high frequency noise suppression processing.Through stabilizing non-stationary seismic signals,empirical mode decomposition gets a series of intrinsic mode functions respective of different frequency bands,where,the high frequency ones often represent high frequency noise.To adapt the space varying of seismic data,this method isolate high frequency data containing part of effective signal from the original data by using global high-pass filtering,converting it to the f-x domain,and then in the sliding short window,extracting the complex modal function IMF1 with high frequency of space varying data sequence in each frequency,finally,after filtering,it can reconstruct to get the time signal after suppressing high frequency noise.The actual data processing results show that,the above three kinds of time-frequency analysis methods,with frequency characteristics themselves,can separate high frequency noise in time and space varying seismic signal and suppress reasonably,after suppressing,the noise profile don't contain valid signal energy,and after reconstruction,the seismic signal is with high signal fidelity,moreover,the signal to noise ratio of seismic profile is significantly improved.