Attenuation of Residual Multiples and Coherent Noise in the Wavelet Transform Domain

Tamara Pokrovskaia and Richard Wombell
Veritas DGC, Crawley, West Sussex, United Kingdom

Although generally very powerful, noise and multiple attenuation techniques can often leave remnants in seismic data. For example, noise from extraneous sources such as rigs and other boats can be hard to model and fully remove using standard methods. Similarly, in areas of relatively complex geology, remnant multiple may sometimes be left after multiple attenuation methods such as SRME and Radon de:-multiple. These remnants can cause problems in later processing, for example through the generation of migration noise and contamination of AVO analysis etc. and therefore often need to be further attenuated in the processing sequence.

An approach to attenuating such residual noise and multiple energy is to exploit the wavelet transform domain, where data are represented in both frequency and time simultaneously, whereas standard noise attenuation methods typically analyze data in either just the time or frequency domain. The Wavelet Transform (WT) was introduced for the analysis of non-stationary seismic signals as an alternative to the Fourier Transform. Because of time-frequency localization of the WT, the separation of signal and noise can be more distinct in different voices (or scales). Therefore application of conventional techniques (such as despike) in the wavelet domain, when applied on selected voices, may be more successful than in time domain or frequency domain alone. The advantage of processing in the wavelet domain is that some spectral components of the signal are preserved untouched, and therefore primaries are less affected. Since it represents data using wavelet functions of finite time duration and frequency coverage, the WT provides a convenient domain in which to estimate signals according to their temporal and spectral characteristics simultaneously.

We show two examples to illustrate the effectiveness of such an approach: the attenuation of residual multiples on deep-water Nigerian data and attenuation of boat noise.