Seismic noise removal

Patent metadata

Field	Value
Publication number	US-10288754-B2
Application number	US-201314389349-A
Country	US
Kind code	B2
Filing date	Mar 29, 2013
Priority date	Mar 29, 2012
Publication date	May 14, 2019
Grant date	May 14, 2019

Abstract

A method for removing seismic noise from an input seismic trace. The method may receive the input seismic trace. The method may receive one or more noise references for the input seismic trace. The method may receive one or more filters corresponding to the noise references. The method may apply a nonlinear function to the input seismic trace and to the one or more noise references to produce respective output signals for the input seismic trace and for the one or more noise references. The nonlinear function may be capable of determining higher-order statistics. The method may update the filters based on increasing one or more information attributes of the output signals to a predetermined threshold. The method may then filter noise corresponding to the noise references.

First claim

What is claimed is:

1. A computer-implemented method for removing seismic noise from an input seismic trace, comprising:
receiving the input seismic trace, wherein the input seismic trace depicts a structure of a subsurface volume;
receiving one or more noise references for the input seismic trace;
determining independent components from the input seismic trace, wherein determining the independent components comprises:
estimating a de-mixing matrix for the input seismic trace using an independent component analysis (ICA) algorithm; and
applying the de-mixing matrix to the input seismic trace to obtain the independent components;

determining one or more non-similar independent components using a similarity criterion by comparing the independent components to the one or more noise references, wherein determining the one or more non-similar independent components comprises:
determining an amount of correlation between a first independent component of the independent components from the input seismic trace and at least one of the one or more noise references is above a predetermined threshold;
in response, identifying the first independent component as similar component;
determining an amount of correlation between a second independent component of the independent components from the input seismic trace and each of the one or more noise references is below the predetermined threshold; and
in response, identifying the second independent component as a non-similar component; and

combining the non-similar independent components to produce a filtered seismic trace, wherein the filtered seismic trace depicts the subsurface volume, and the filtered seismic trace contains less noise than the input seismic trace.

2. The method of claim 1 , wherein the ICA algorithm is based in part on information attributes comprising mutual information between the input seismic trace and each of the noise references.

3. The method of claim 1 , wherein the ICA algorithm is based in part on information attributes comprising joint entropy between the input seismic trace and each of the noise references.

4. The method of claim 1 , wherein comparing the independent components to the one or more noise references comprises comparing each of the independent components individually to each of the one or more noise references.

5. The method of claim 1 , wherein the de-mixing matrix comprises a deconvolution matrix.

6. The method of claim 1 , wherein combining the non-similar independent components is performed using an independent component analysis mixture matrix.

7. The method of claim 1 , wherein the ICA algorithm comprises a FastICA algorithm.

8. The method of claim 1 , wherein estimating the de-mixing matrix comprises maximizing a mutual information or joint entropy value of an objective function to determine the de-mixing matrix.

9. The method of claim 1 , further comprising subtracting the similar components from the input seismic trace one-by-one using an adaptive deflation process, wherein the adaptive deflation process includes applying a term to the independent components that is determined by minimizing an energy of the filtered seismic trace.

Assignees

• Westerngeco Llc

Inventors

• Liu Kuang-Hung
• Dragoset Jr William H

Classifications

• G01V2210/48

  Other transforms · CPC title

• G01V2210/3246

  Coherent noise, e.g. spatially coherent or predictable · CPC title

• G01V1/364 Primary

  Seismic filtering (G01V1/37 takes precedence) · CPC title

• G01V1/36

  Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy · CPC title

• G01V1/247

  Digital recording of seismic data, e.g. in acquisition units or nodes · CPC title