Composite Denoising Method Research Based On Multi-domain Transform

At present, the oil and gas exploration gets into the new times.The accuracy of theseismic data is also required higher and higher.In other words，"high signal-to-noise ratio","high resolution","high fidelity" are put forward higher expectations. So improvingsignal-to-noise ratio is the primary task and it directly affects and decides the resolution andthe fidelity.Without the guarantee of high SNR，advanced processing technology and methodsis vain. Since denoising has also been throughout the seismic data processing, manyscholars invest much time in the research of improving signal-to-noise ratio.Now,Wavelet transform and Curvelet transform denoising, is very popular in the area ofnoise suppression. Based on the Wavelet domain and Curvelet domain denoising, highsignal-to-noise ratio data denoising can get good noise suppression effect. But now with thedepth of the exploration, the surface and underground geological effects, the collected seismicdata is often shown as low signal-to-noise ratio and weak signal. Conventional Wavelettransform and Curvelet transform denoising can remove much random noise,but meanwhilewill also cause losing much information of seismic event, reducing the quality of seismicdata.So now,on the Wavelet domain and Curvelet domain denoising method of research, mostare based on two kinds of improvementt to remove noise as much as possiple at the same timeto maintain effective signal on the least damage. A routine is to modifijy hard and softthreshold.Another way is to combine Curvelet transform with statistical theory or optimizedchoice theory to form a composite denoising method.So this paper also use this two way toimprove the traditional Curvelet transform to achive the purpose of amplitude-preserveddenoising.In this paper, through the model and actual data processing,we compare theconventional threshold of Wavelet transform and Curvelet transform denoising effect and check the superiority and the insufficiency of both method.To the insufficient of conventionalCurvelet transform denoising, we combine with GCV (generalized cross-validation) rules andgenetic algorithm, to form a new composite denoising method. This method is trying to usegenetic algorithm to seek a more ideal threshold after many iteration. At the same time, thismethod need not know the signal-to-noise ratio of the a priori information in advance.So it’smore suitable for practical application. Through the model and actual data test, improvedCurvelet transform denoising, contrasted with the traditional Curvelet transform denoisingmethod, is more useful in improving signal-to-noise ratio and can get further fidelity.