Study On Seismic Resolution Enhancement Methods In Time-frequency Domain

Enhancing resolution of seismic records is one of the important step during seismic data processing, which play an important role on the subsequent processing and interpretation. The assumption of traditional enhancing resolution method that the seismic signal is stationary does not meet the actual situation, therefore, it is difficult to obtain good results. During the process of field acquisition, because of the influence of such factors as acquisition environment, terrain, instrument and so on, the acquired data may have some problems such as missing and bad traces, as well as irregular data distribution, which leads to spatial aliasing becoming more serious after the processing of improving resolution. In addition, the acquired data with random noise including environmental noise and the existence of regular interference wave leads the signal-to- noise ratio(SNR) of sections after improving the resolution is lower, which brings new chal enges to the processing of improving resolution.Traditional method of improving resolution is often carried out in single trace. Because of the differences of traces in seismic records, it certainly leads the lateral continuity of the records becoming worse after processing. Traditional single-trace deconvolution model is also inefficient for the recovery of missing traces. Curvelet basis function has the characteristics of scale and direction and also is supported by certain time-space domain. In this paper, the improving resolution of irregular data in the noise environment is introduced into the C urvelet domain, and the regularized sections with high SNR and resolution are finally obtained by sparse promoting solution, which serves for the subsequent processing and interpretation work of thin layer oil reservoir.This paper focuses on the research of improving resolution met hod in S domain and Curvelet domain. The research contents mainly include: 1) Aiming at the problem of extracting the high-order statistics mixed phase wavelet, we take the CMP gathers in seismic records as one system with single input and multiple outputs and make use of the system phase information included in the second-order cyclostational statistics to deduce and construct wavelet estimation expression, and thus realize the method of extracting mixed phase wavelet based on the SIMO system. 2) As seismic wavelet often changes with time, in order to extract time- varying seismic wavelet, we extend and develop the traditional spectral modeling method to S domain based on the excellent time- frequency local characteristics of S transform. And in the process o f constructing time- varying wavelet estimation equation, as the influence of local energy attenuation abnormity caused by the strata bearing oil and gas on the wavelet extraction is considerate, the abnormity are not destroyed or eliminated after the processing of improving resolution. 3) Aiming at the problem that the efficiency of the spectral modeling method in S domain is low, we analyzed the differences between the characteristics of time-varying wavelet and reflection coefficient in time- frequency secondary spectrum. The rapid extraction of time-varying wavelet spectrum is realized by setting up reasonable filter. 4) Making use of the sparse representation characteristic of C urvelet transform, in combination with the convolution operator generated by the extracted wavelet, we transform the improving resolution into sparse promoting matching inverse in the C urvelet domain. The ‘mask’ operator is introduced to realize data regularization as well as the resolution improvement; 5) The relationship between the geometric characteristics in time-space domain and the phase spectrum of signal is analyzed. The transformation strategy of phase spectrum is proposed in order to make full use of the phase information of signal. And the operator separating signal and noise is designed, which realizes the target of improving resolution as well as separating linear noise in the Curvelet domain.The numerical experiments and real dada tests in the paper verified the validity of the theoretical method. The processing results show that the bandwidth of the data frequency becomes wider after improving resolution in the time- frequency domain, the dominant frequency is improved and the missing seismic data is also recovered, which restrains the noise reasonably and effectively, relieves the conflict between resolution and SNR, and further develops the conventional method of improving resolution.