Research On High-density Seismic Weak Signal Detection And Denoising With Curvelet Transform

High-density seismic technology was one of the geophysical exploration technologies which had developed rapidly in recent years. The high-density seismic data had solved the problems such as suppressing noise, enhancing resolution and keeping fidelity. Meanwhile there were also some critical problems, such as low SNR and sophistication weak signals with noise. The quantity of the final data wouldn't be perfect if we used the original data without denoising for stacking, velocity analyzing, NMO correcting and migrating. It's difficult to exert the potential advantages of high-density data if we used the original data without denoising. Therefore, it's of great importance to research on weak signal detection and denoising of high-density seismic wave and find out how to improve SNR of high-density seismic data.Curvelet transform was a new multi-scale transformation which developed on the basis of wavelet transform. Curvelet transform worked better in display the edge of graphics, such as curves, straight lines and other geometric features, this feature made Curvelet transform achieve abundant research results in investigation. Curvelet transform was introduced into the field of density seismic technology in this paper. The usage of Curvelet transformation in high density seismic weak signal detection and denoising was emphasized discussed.In this paper, Fast Discrete Curvelet Transform was used as a basic tool to discuss how to deal with seismic data contains weak signal or not. In case of no weak signal in seismic data, denoising with hard threshold value method may achieve good effect. While weak signal was contained in the data, lots of threshold value methods were used to evaluate in weak signal identification separately. The semi-soft threshold method was chosen which had better effect than hard threshold and other methods, and the effect of the weak signal identification was improved through reducing the threshold.To overcome the problem of filtering coefficients thoroughly in Curvelet domain and singular points in time domain, the mean filter method was used in Curvelet domain. Model and real data tests verified that this process achieved better effects than other methods, such as Wavelet transform and hard threshold method in Curvelet transform. This algorithm could discern the weak signal buried under 2 times the noise signal amplitude.