Seismic deconvolution based on SNR-weighted stacking

We propose a seismic deconvolution algorithm based on signal-to-noise ratio (SNR)-weighted stacking. First, we applied the time-frequency analysis to decompose the common midpoint (CMP) or common reflection point (CRP) gather. It is considered that there are differences in noise between different traces in the frequency domain profile. Then, we employ an SNR-weighted stacking strategy in each frequency-domain profile. Finally, we construct the resolution-enhancement amplitude spectrum, apply a linear inversion approach to calculate the overall weights of each frequency-domain profile, perform weighted stacking between the frequency-domain profiles, and obtain the resolution-enhancement post-stack data. The frequency decomposition of the pre-stack gather is efficiently utilized for the primary data of the proposed method. The SNR-weighted stacking adaptively selects the components with high SNR at each frequency. In addition, the resolution-oriented stacking weights are determined by the linear inversion approach. This algorithm can be adaptively weighted by considering the SNR of each frequency component, which can effectively overcome the problem of low SNR in spectral simulation and obtain high-resolution seismic data. The algorithm does not include wavelet extraction and regularization; thus, the resolution-enhancement data is more reliable. Applications to synthetic and field seismic datasets demonstrate that the proposed resolution-enhancement methodology is of great benefit to the succeeding structural interpretation and thin-bed identification.