About This Item

Share This Item

The AAPG/Datapages Combined Publications Database

AAPG Bulletin

Abstract


Volume: 66 (1982)

Issue: 10. (October)

First Page: 1683

Last Page: 1684

Title: Application of Nonlinear Constraints to Processing of Seismic Data: ABSTRACT

Author(s): F. Baixas, P. Hooyman, C. Wu, R. Inguva, L. Schock

Article Type: Meeting abstract

Abstract:

End_Page 1683------------------------------

Astronomers Gull and Daniell have developed efficient noise-reduction techniques to enhance radio galaxy displays. They used the information theory notion of entropy with a chi-square constraint. In these conditions, the criterion derivatives provide separate equations to estimate each processed sample at a very fast computation rate. The chi-square constraint serves as a global constraint and ensures the continuity of the process over a given area and yields a confidence level for the results. The full confidence situation corresponds to the standard stack, whereas the zero confidence level corresponds to a flat distribution (trace). By slightly decreasing the confidence level from 100% confidence, meaningful noise reduction is achieved.

This technique is applicable to the noise reduction of seismic data. At each time sample of a deconvolved gather, the mean and the standard deviation are associated within the chi-square constraint to generate a new stack estimator. Such a process may be effective in sharpening velocity analyses. This is shown on both synthetic and real data. This nonlinear constraint method can be advantageously compared with more classical semblance or coherency technique with respect to computer time and the number of selected parameters is minimum.

Bryan and Skilling tried a new statistic in order to obtain a better distribution of noise residuals. The proposed solution consists of constraining the standardized and ordered residual components of the chi-square to fit a theoretical normal distribution. In fact, the solution can be made more general according to the noise distribution, inferred experimentally from the data.

Attempts have been made to apply this technique to the same seismic data sets. The Gull and Daniell technique provided a first estimate, which served as a starting point for the second technique. Convergence was obtained after a few iterations and improvement of the solution with respect to a normal distribution was checked through a chi-square goodness-of-fit procedure.

End_of_Article - Last_Page 1684------------

Copyright 1997 American Association of Petroleum Geologists