Abstract
Marin seismic acquisition results in signals that are affected by both random and coherent noise. For successful imaging it is important that as much as possible of this noise is attenuated in an early stage of processing. One of the effective methods to handle random noise, such as weather noise, is a time frequency de-noising (TFDN) algorithm. In this thesis first a short introduction is given to the different classes of noise that can contaminate seismic data, with emphasis on weather noise. Also a brief overview of existing commercially used de-nosing methods for random noise attenuation will be presented. The main objective of this thesis has been to investigate an existing TFDN algorithm, and search for improvements. Several new algorithms for locating and attenuating noise is described and tested. The various algorithms have been compared employing a controlled data set, and using an industry standard TFDN program as benchmark. Based on the results obtained, a selection of the most promising algorithms was further analyzed employing real data contaminated with weather noise. The overall conclusion is that the industry standard TFDN program can be further improved and possibly replaced by more efficient de-noising algorithms.