| To meet the increasing demand of energy, seismic exploration has been shifted from the east to the northwest and from plains to deserts, where the seismic data gathered are always contaminated by severe linear noise and surface wave. With the presence of surface wave, the signal-to-noise ratio of the seismic data decreases rapidly, and the validness of the seismic data has been gravely damaged. Therefore, the surface wave in the pre-stack must be suppressed, while the conventional denoising methods, such as f-x filtering, f-k filtering, wavelet transform and so on, perform ineffectively to suppress surface wave.In this paper, we first introduce the characteristics of the surface wave, and the basic theory and properties of multi-scale analysis. Then, based on the differences of spectrum between surface wave and the reflection signal, the seismic data has been sorted into different bands using the 2-D wavelet transform and Curvelet transform respectively, both of which are able to achieve multi-scale transform. With regard to the protection of signals in other bands, the processing has been applied to band dominated by surface wave. The main contents of the paper are as follows:First,2-D wavelet transform was exploited to treat seismic record. The dominant frequency band of the surface wave in the frequency-wave number domain was determined prior to other treatments. Meanwhile, a practical connection was established between the selection of thresholds and methods of constructing thresholds in image processing, which paved the way for the introduction of average threshold and histogram threshold. In the second place, wavelet coefficients of the determined band of surface wave were further segregated with the two proposed image thresholds. These two image thresholds and two other common threshold functions were applied to theoretical analysis and actual seismic data respectively and results showed that the signal-to-noise ratio and the resolution of the seismic data were improved remarkably through the processing of histogram threshold. In the last place, the algorithm posed in this paper based on the wavelet transform and histogram threshold, together with conventional methods, was utilized to suppress surface wave in actual seismic data, which demonstrated that the proposed algorithm effectively suppressed surface wave and enhanced the reflection signal.Then, Curvelet transform was implemented for the processing of seismic data, which has been proposed as an alternative to wavelet transform on the basis of wavelet analysis and is equipped with the ability to best represent two-dimensional signal with some plane curve singularities. The critical steps of Curvelet transform are specified as follows:Step 1. Seismic data was processed in the utilization of Curvelet transform to determine the leading band of surface wave in the Curvelet domain; Step 2. A corresponding relationship between the time domain and the direction of Curvelet domain was advanced for the purpose of defining Curvelet coefficient matrix in various orientations just according to the angle ranges of surface wave, which takes advantages of low apparent velocity of surface wave; Step 3. The Curvelet coefficient matrix was iterated through the proposed threshold derived from weighted average and other three thresholds. Compared with the other three thresholds, both theoretical model analysis and processing of actual seismic data demonstrated the effectiveness of the improved threshold.Finally, the two proposed algorithms are applied to actual suppression of surface wave in seismic data respectively, and comparative analysis of results proved that Curvelet transform performed more effectively to maintain the continuity of the reflection signal than wavelets. |
