Research On De-Noising Methods Of Seismic Data

To improve the resolution of seismic data is a main task of geophysical exploration, and its prerequisite condition is to improve the signal-to-noise ratio. It is necessary to remove coherent noises and random noises in seismic data processing to improve the signal-to-noise ratio. Commonly there are some rules of coherent noises, so they can be removed according to their rules. Random noises have no rules, so it is difficult to remove random noises. Therefore how to effectively get rid of random noises is a goal which many people want to reach. Research on how to get rid of random noises effectively is the main work of this paper, and how to get rid of linear coherent noises is also concerned. In this paper, other techniques are discussed, such as wavelet transform and KL transform.Aiming at characteristics of noises of seismic data, several methods of removing coherent noises and random noises are briefly introduced, and their strongpoint and shortcomings are analyzed too. Then wavelet transform is introduced in detail.At the beginning of the third chapter, conventional de-noising method by threshold filter in wavelet domain is introduced, and for the sake of overcoming it's limitations, a new method which is named de-noising method by threshold filter in wavelet domain based on quadratic wavelet transform is proposed. Comparing with the conventional method, the new method has better de-noising effect, and can hold the main details of seismic data well. This new method also can be used for dealing with other signals except for seismic data.To improve on the conventional de-noising method by threshold filter in wavelet domain, a new method which is named de-noising method by combines time-frequency correlation analysis with threshold filter in wavelet domain is proposed. By use of this new method, not only the different features of signals and noises in wavelet transform are used for removing noises, but also the correlations of the seismic paths are used. The results show that the new method can improve the effect of disposal and the signal-to-noise ratio by processing the synthetic data and real data.Wavelet transform modules maxima can be used for removing noises in seismic data processing, Compare with other methods, this method can accurately distinguish signals from noises in different scales, and the high frequency noises can be greatly removed, while the high frequency signals can be remained as much as possible. The limitation of this method is discussed in fifths chapter, and the direction of research is pointed out.Aiming at limitations of conventional KL transform for removing noises in seismic data processing, a new method which is named space-time variety and dip scanning stack KL transform is proposed in this paper. This method is fit for dealing with those sections which contain oblique or curving events, and can dispose these sections neatly. By use of this method the whole section can be plotted into several fields, which field can include different number of paths, can have different beginning time and terminate time. Then the paper points out what should be pay attention to in using this new method. Another new method which is named space-time variety and dip angle KL transform is proposed, this method is fit for removing linear strong energy coherent noises. By use of this technique, not only the strong coherent noises could be removed effectively, but also thecontinuity of effective events could be kept well. Furthermore, these two methods could be used together. When there is strong noises in seismic data, conventional KL transform can't effectively remove noises, nonlinear KL transform processing is a good choice for seismic data. Another de-noising method, proposed in this paper, a combination of KL transform and quadratic wavelet transform is also a good choice. This method is particularly suitable for seismic data which include strong energy noises. It can improve the signal-to-noise ratio in seismic data well. In the final chapter, conclusion and expectation are given.