Multidimensional techniques of signal processing applied to seismic data

Zusammenfassung

This Ph.D. thesis dissertation addresses seismic data processing techniques in the active seismic experiment context, a key issue in the study of the Earth's structure and of the physical properties of the materials that form the Earth. The seismic data processing is an area where have been traditionally applied the most advanced processing tools to overcome problems of filtering, separation and monitoring of seismic waves with the most demanding specifications. This dissertation thesis examines techniques of analysis in the time-frequency/time-scale domain and of filtering in slowness (inverse of velocity) most commonly used in geophysics. The latter are better known in other fields as directional filters. Starting in this context, new slowness adaptive filters are developed combining the windowed Radon transform (local slant-stack) with the estimators of the instantaneous slowness of the wavefronts based on measurements of the degree of coherence. These filters are then extended combining the wavelet transform and the windowed Radon transform to build a new type of directional filters called adaptive slowness filters in the time-scale domain. In both cases, a special attention is given to provide objective criteria to allow the selection of the optimal parameters for each application, and the performance of the filters is compared to the most commonly used techniques. With these new tools, a wide variety of slowness adaptive filters can be designed to make the most of the high lateral coherence of the seismic waves and to process the seismic record sections in a simple and flexible way. The two main applications of these tools are to automatically separate or filter the seismic waves from complex seismic record sections without having an accurate measure of their slownesses, and to distinguish the low-energy wavefields from the background noise and from the other seismic waves. After some objective tests on a wide range of synthetic data, the adaptive slowness filters in the time and time-scale domains have been employed with great success on several real record sections from large marine seismic profiles of wide-angle experiments to smaller, but not less complex, ground seismic profiles. In the wide-angle seismic profiles, the refracted waves have been successfully separated from the background noise and other interfering signals, and high-energy water waves have been attenuated causing minimal distortion to the other signals, and in particular to the low-energy refracted waves under study. In the ground seismic profiles, the high-energy surface waves, such as the Rayleigh waves, have been separated to see the P (pressure) and S (shear) waves of interest.