Abstract
Improvements in seismic resolution beyond typical seismic wavelength will have significant implications for hydrocarbon exploration and production. Conventional imaging algorithms can be derived as a least squared optimization problem in which the resulting algorithm is a cross-correlation (second order statistics) operation whose region of support is limited to the bandwidth of the source signal. This is not the case for non-vanishing higher order cumulates where the support region can be extended beyond the typical seismic wavelength. For cases of non-vanishing higher-order cumulate, we have reformulated the present seismic imaging operator (cross-correlation) to look like a bicoherence-correlation operator (third order statistics). Numerical examples with data corresponding to heterogenities smaller than the typical seismic wavelength and containing non-Gaussian noise confirm that higher resolution can be obtained by using our algorithm over conventional imaging algorithms. We also present a bicoherence-correlation based automatic time picking algorithm which is used for detecting events or relevant patterns in seismic data which is very crucial for data interpretation and processing. This algorithm is not just limited to the first break picking (picking of events that arrive first), as is the case for conventional automatic time picking algorithms based on cross-correlation method and neural networks. The algorithm can also be used for autographing of a desired event by specifying the start time of the event we want to track.
Srinivasan, Karthik (1999). Seismic imaging using higher order statistics. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -1999 -THESIS -S70.