Summary
Recognition and identification of effective signals with low signal-to-noise ratio (SNR) is one of the difficulties for seismic data processing. The super-virtual interferometry (SVI) method is able to promote SNR of interferometric signals which satisfy stationarity conditions, and achieves good results of building velocity models and diffraction imaging with low SNR data. Either the correlation or convolution based SVI method is susceptible to additive noise and is only discussed in 2D cases. To overcome these limitations, this paper develops the higher-order cumulant based coherent integration (HOCCI) method to enhance interferometric signals by substituting the process of correlation and convolution in the SVI method by higherorder cumulant and multidimensional convolution. The 3D synthetic data examples demonstrate that the HOCCI method, compared with the SVI method, has better performance in promoting SNR of interferometric signals and suppressing coherent noise, and the 3D field data examples further confirm the accuracy of the 3D scheme of HOCCI method.
Introduction
The super-virtual interferometry (SVI) method can enhance interferometric signals satisfying stationarity conditions, which partly share the common raypaths (Snieder et al, 2006). The commonly discussed interferometric signals include refractions from the same layer and the diffracted waves from the same diffraction point. The SVI method is implemented by two steps: correlation and summation of seismic data to generate virtual traces with enhanced interferometric signals, followed by convolution with actual traces to obtain super-virtual traces with further enhanced interferometric signals (Mallinson et al., 2011; Bharadwaj et al., 2012). This method achieves good results of extracting diffracted waves to image deep structures (Dai et al., 2011) and of promoting the SNR of first breaks in low SNR traces for undulate surface areas (An et al., 2014).
However, the SVI method has some limitations. (1) The irregular geometry, which is designed to adapt to undulate surface conditions, may lead to few available sources and receivers which generate interferometric signals, such that the SNR enhancement of the SVI method is limited in this case. (2) The process of correlation in the SVI method is susceptible to additive noise, especially when additive noise in different traces is coherent. (3) The SVI method is only discussed in 2D cases and cannot be completely implemented in 3D cases.
