The Research And Application Of Deconvolution Method Based On Gabor Transform

In the seismic exploration, along with the high-resolution demands, the high-resolution seismic processing has been widely cognized and adopted. But there are a lot of constraint conditions in conventional deconvolution methods in the high resolution data processing. Consequently, it can't improve the resolution effectively, and its effects cannot satisfy the geophysicist's requirement.Usually, there are three important factors which must be considered in the deconvolution of seismic data: minimum phase of seismic wavelet, white noise of reflection coefficient, and ideal expected wavelet. Seismic wavelet is a non-minimum phase, we can make it minimum phase by doing exponential weighting in the conventional deconvolution processing. As for nonwhite noise problem of reflection coefficient, traditional statistical wavelet deconvolution is using the statistical average of the autocorrelation function of seismic data in common shot point and common receiver domain. Whereas, the seismic wavelet estimated with this method still don't overcome the effects of nonwhite noise problem.; In addition, expected wavelet is bandlimated wavelet or ricker wavelet in traditional statistical wavelet deconvolution methods. Regarding a bandlimated wavelet, The former has long delay period and complicated side-lobe waveform, and the latter has high side-lobe amplitude. So we have to find a better method to seek for the ideal expected wavelet.This thesis is proceeding from seismic deconvolution model direct problem, and expounds the basic concept and approach of deconvolution in detail. Then the paper introduce the theory and method of deconvolution in conventional seismic data processing,such as wiener deconvolution, spiking deconvolution, predict deconvolution, homomorphic deconvolution and conventional surface-consistent deconvolution.We give realization steps of predict deconvolution and application concept of conventional suface-consistent deconvolution.Generally speaking,conventional deconvolution method is mainly applied into fourier transform.,while fourier transform is a form of processing the signal globally. This processing is very good at continuous stablly signal.But fourier transform is not very good at nonstationary signal or distorted signals with formation reflection. To overcome the shortage of anslysis nonstationary signal with fourier transform, we are using a special fourier transform, it is called Gabor transform. Using a localization window function g ( t ? b)draw into the fourier transform, parameter is moving along the time domain.That is to say, adding windows on time direction and frequency direction in conventional fourier transfrom, this processing considers not only the change in time domian ,but also the attenuation in frequency domian of seismic signal.This paper is studied according to the developing and foreqround of deconvolution technology on the basis of predecessor's work, combined with practical production, we are carrying on deeply research about Gabor deconvolution. The wavelet is a attenuation signal beause in real seismic data the fomation has resorptive action on wavelet . Different from conventional deconvolution, Gabor deconvolution don't try to pick up the wavelet, it just estimates the wavelet considering its nonstationarity attenuation. Assumed that the formationreflection coefficient is white noise sequence, the wavelet transmiting in the formation emerges a decay seismic trace, and then Gabor transform is used to carry out spectrum analysis on that trace, dealing with decay seismic trace using smoothing process to remove the impact of the reflection coefficient, finally we get the amplitude spectrum of wavelet.To take a step further, assumed that seismic wavelet is minimum phase, through the Hilbert tranform we obtain the phase of the wavelet,and finally implemnet the estimate of decay wavelet. Calculation results from theoretical models and practice processing indicate that the resolution of seismic profile processed with Gabor deconvolution is improved, and the frequency band is broadened, so the spectrum is compensted. Gabor deconvolution not only suits pre-stack data,but also suits stacked data.We try to extend Gabor deconvolution to the full scope Gabor surface-consistent deconvolution.This method is not to pick up the windows for the seismic data respectively, it just carry out surface-consistent deconvolution on wavelet. And then Gabor deconvolution becomes a part of inhibit amplitude processing. The main processing is to make Gabor transform for every trace,and then smooth the data to get the estimate of the wavelet. We can gain the estimate factor of the CDP traces and do geometric mean for the every estimate factor to get estimate of the attenuate wavelet. The receiver component is the same as the source component.Finally, we will get reflection coefficient series.Gabor deconvolution has been applied to field data processing, and got a quite satisfied results, which broden the frequency-bandwidth and compensated the amplitude. In gabor domain, the combination of surface consistent and time-difference broadens frequency spectrum and stabilizes side-lobe of wavelet. In comparison with the conventional surface-consistent deconvolution methods, the Gabor surface-consistent deconvolution has the superiority on freqency-spectrum compensation and lateral continuity. The algorithms of Gabor surface-consistent deconvolution has the characteristics of denoise, phase adaptation, amplitude enhance and thin-bed recognization ability, which overcomes the overcompensation of frequency-components and nonuniformity in different time-windows within the same wave group in spatial domain. The potentials of Gabor deconvolution and Gabor surface-consistent deconvolution lie in improving seismic signal's resolution, especially on records of high frequency components and deep layers'.