Research On Pre-stack Near Offset Data Reconstruction And Surface Related Multiple Elimination

Multiple elimination has always been a tough problem in marine seismic data processing. Multiple wave disturbs the identifiability of primary reflection, causing false seismic imaging which may result in wrong seismic interpretation and eventually may cause serious economic losses. Existing filter based methods, which are used to attenuate multiple of real data, are not only limited in some way in practical use but also harm primary. Technology of surface related multiple elimination (SRME) based on wave equation does not require geological data of the subsurface, and is widely used in production. However, the SRME is limited to certain conditions and the cost for the 3-D SRME is still relatively high. Therefore, the research prospective and direction of the development is to improve the algorithm and applicability of the SRME method.The present paper first reviews the theories of wave-field extrapolation, matrix of WRW model and feed back model, intending to provide a deep understanding of the technology of SRME, and then explains the applicability and limitations of the method. After that, the deduction of focal transform from wave-field extrapolation and feed-back model theory is explicated. The method of focal transform, originated from the SRME method, uses cross-correlation process instead of original weighted convolution. Thus focal transform is exactly the opposite of the SRME. The present paper applied focal transform to model-drived focal operator and the original data of the lost near offset data. Reconstructed near offset data can be gotten from the matching subtract of removed focal point data and the raw data. Unlike other data reconstruction methods, this method is complete data drive with no need for any subsurface geological data, adaptable to any complex structure, and has better amplitude-preserved.Based on feed-back model, the relationship between multiple and primary in the inverse data domain is deduced. In the inverse data domain, primary and surface related operator can be separated. Through adaptive removing of the focal operator, the surface related multiple will be eliminated. In the present paper, QR decomposition is firstly used to solve the complex matrix inversion. The QR decomposition method can improve the stability and precision of the inverse transform. The inverse data domain multiple elimination can be seen as the transformation and improvement of the SRME technology and the expression is very simple and explicit, in which multiple can be better separated from primary in inverse domain and does not cause any damaged effect on primary energy. The synthetic and real examples clearly justify the validity of the technology presented in this dissertation.