Stacking seismic data using local correlation

Introduction

Stacking as one of the three crucial techniques (deconvolution, stacking, and migration) plays an important role in improving signal-to-noise ratio (S/R) in seismic data processing (Yilmaz, 2001). Conventional stacking, which is performed by averaging an NMO-corrected data set or migrated data set, is optimal only when noise components in all traces are uncorrelated, normally distributed, stationary, and of equal magnitude (Neelamani et al., 2006; Mayne, 1962). Therefore, different stacking technologies have been proposed, along with improvements in optimizing weights of seismic traces.

Robinson (1970) proposes using an S/N-based weighted stack to further minimize noise. Using cross-correlation of seismic traces and normalized cross-correlation processing, Chang et al. (1996) proposes preserved frequency stacking. Schoenberger (1996) proposes optimum weighted stack for multiple suppression, with weight determined by solving a set of optimization equations. Neelamani et al. (2006) propose a stack-and-denoise method called SAD, which exploits the structure of seismic signals to obtain an enhanced stack. Zhang and Xu (2006) present a high-order correlative weighted stacking technique on the basis of wavelet transformation and high-order statistics. By estimating the probability distribution of noise, Trickett (2007) applies a maximum-likelihood estimator to stacking. To eliminate artifacts in angle-domain common-image gathers (CIGs) caused by sparsely sampled wavefields, Tang (2007) presents a selective stacking approach that applies local smoothing of envelope function to achieve the weighting function. Rashed (2008) proposes smart stacking, which is based on optimizing seismic amplitudes of the stacked signal by excluding harmful samples from the stack and applying larger weight to the central part of the sample population.

The global correlation coefficient can measure the similarity of two signals, but it is not a local attribute. Not only does the sliding-window global-correlation approach need many parameters to be specified, but this approach cannot smoothly characterize thin layers well. Fomel (2007b) uses shaping regularization, which controls locality and smoothness to define local correlation (Fomel, 2007a). Local correlation is applied to multicomponent seismic image registration (Fomel et al., 2005; Fomel, 2007a) and time-lapse image registration (Fomel and Jin, 2007).

In this paper,we present a new stacking method using local correlation. This method applies time-dependent smooth weights (which are taken as local correlation coefficients between reference traces and prestack traces), stacks the common-midpoint (CMP) gather, and effectively discards parts of the data that least contribute to stacked reflection signals. Using synthetic and field data examples, we show that, compared with other stacking methods, this method can greatly improve the S/N and suppress artifacts.

Stacking seismic data using local correlation