[First Hit]

Datapages, Inc.Print this page

Field Data Performance of Curvelet-Based Noise Attenuation

Ramesh Neelamani1*, Anatoly Baumstein1, Mohamed Hadidi2, and Warren Ross1
1ExxonMobil, USA
*[email protected]

This study presents the efficacy of curvelets, a recently developed Previous HitmathematicalNext Hit transform, in attenuating random and coherent linear noises in a stacked dataset from the Middle East. Our main motivation to seek a new, advanced noise-attenuation tool is that even though conventional filtering techniques such as median filtering and FX-Previous HitdeconvolutionNext Hit remove respectable amounts of noise, they also harm the signal. The curvelet transform is a recently developed Previous HitmathematicalTop tool that represents an image using a linear, weighted combination of special elementary functions that resemble small pieces of a band-limited seismic reflector (Candes and Donoho, 1999). Each curvelet elementary function has a characteristic dip, frequency (thickness), and location. The localized nature of curvelet functions, along with their dip and frequency characteristics, makes the curvelet transform particularly suitable for attenuating noises in seismic data. In seismic data, most noises differ from the underlying geological signal in terms of the dip, frequency, and/or location. Consequently, signal and noise separate more effectively in the curvelet transform domain than in other conventional transform domains that do not simultaneously exploit all these attributes. This powerful property enables us to separate the geologic signal from the noise by carefully muting appropriate curvelet components of noisy data. Our results demonstrate that the curvelet-based approach provides superior noise attenuation, with minimal impact on the desirable signal components. In conjunction with the preceding multiple attenuation steps that were employed on the dataset (see presentation by Baumstein et al for details), the noise suppression significantly improved the structural and quantitative interpretability of the dataset, thereby validating the efficacy of our approach.


AAPG Search and Discovery Article #90077©2008 GEO 2008 Middle East Conference and Exhibition, Manama, Bahrain