利用分频与反演提高地震资料信噪比与分辨率方法研究

This thesis mainly talks about the wavelet transfrom and the frequency division method. It describes the frequency division processing on prestack or post-stack seismic data and application of inversion noise attenuation, frequency division residual static correction and high resolution data in reservoir inversion. This thesis not only describes the frequency division and inversion in theory, but also proves it by model calculation. All the methods are integrated together. The actual data processing demonstrates the applying results. This thesis analyzes the differences and limitation between t-x prediction filter and f-x prediction filter noise attenuation from wavelet transform theory. It considers that we can do the frequency division attenuation process of noise and signal by wavelet frequency division theory according to the differences of noise and signal in phase, amplitude and frequency. By comparison with the f-x coherence noise, removal method, it approves the effects and practicability of frequency division in coherence and random noise isolation. In order to solve the side effects in non-noise area, we: take the area constraint method and only apply the frequency division processing in the noise area. So it can solve the problem of low frequency loss in non-noise area. The residual moveout differences in seismic data processing have a great effect on stack image and resolutions. Different frequency components have different residual moveout differences. The frequency division residual static correction realizes the frequency division and the calculation of residual correction magnitude. It also solves the problems of different residual correction magnitude in different frequency and protects the high frequency information in data. By actual data processing, we can get good results in phase residual moveout differences elimination of pre-stack data, stack image quality and improvement of data resolution. This thesis analyses the characters of the random noises and its descriptions in time domain and frequency domain. Furthermore it gives the inversion prediction solution methods and realizes the frequency division inversion attenuation of the random noise. By the analysis of results of the actual data processing, we show that the noise removed by inversion has its own advantages. By analyzing parameter's about resolution and technology of high resolution data processing, this thesis describes the relations between frequency domain and resolution, parameters about resolution and methods to increase resolution. It also gives the processing flows of the high resolution data; the effect and influence of reservoir inversion caused by high resolution data. Finally it proves the accuracy and precision of the reservoir inversion results. The research results of this thesis reveal that frequency division noise attenuation, frequency residual correction and inversion noise attenuation are effective methods to increase the SNR and resolution of seismic data.