Browse Prior Art Database

ITERATIVE AUTOMATIC GAIN CONTROL OF SEISMIC DATA

IP.com Disclosure Number: IPCOM000240361D
Publication Date: 2015-Jan-27
Document File: 5 page(s) / 600K

Publishing Venue

The IP.com Prior Art Database

Abstract

A novel way to get around problems with determining RMS windows to use is using an iterative RMS operator. So, instead of doing one single iteration of RMS, with a window-size of length n, one can (as an example) instead do three RMS operations with window length n/2.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

ITERATIVE AUTOMATIC GAIN CONTROL OF SEISMIC DATA

Automatic Gain Control (AGC) of seismic data is often done by dividing / normalizing each sample in the cube with a measure of the local amplitude level. The amplitude level is usually calculated through a windowed Root-Mean Square (RMS) process. The window size in the RMS operator determines the dynamic range of the AGC result. A very short RMS operator will give a small dynamic range, as each sample then will contribute relatively much to the RMS estimate. A large RMS window will give higher dynamic range (i.e. better separation between high-amplitude and weak-amplitude reflectors) in the result. The RMS operator is however sensitive to individual strong reflectors, giving the RMS volume a stair-step/boxcar-filter appearance. When we use such a volume as normalization factor, we will introduce sudden gain changes in the AGC output. This is not desirable, as it will introduce some high-frequency noise in the AGC result.

A novel way to get around this will be to do an iterative RMS operator instead. So, instead of doing one single iteration of RMS, with a window-size of length n, one can (as an example) instead do three RMS operations with window length n/2.

We specifically mention RMS operator in this document, but it should be clear that any windowed operator which provide a measure of amplitude in the area close to the sample at hand is covered by our innovation, e.g. iterative Median, Mean, RMS, Envelope, Robust Mean, etc.

To be more precise, we have these alternatives:

a)      AGCa(C,n) = C / RMS(C,n)

b)      AGCb(C,n) = C / RMS(RMS(RMS(C,n/2),n/2),n/2)

Where a) is the classic method and b) is the new and improved method. The symbol C represents the input seismic cube. In this example we do three iterations with a short window instead of one iteration with a larger window. In practice, we want both the number of iterations, and the window length, to be parameter-controlled.

 Here is one example.

 Left: Input seismic

Mid: One RMS iteration with length 21 samples. Note the blocky appearance.

Right: Three RMS iterations with length 11 samples in each iteration. Note the absence of blocking.

Here we see the same RMS result for one single trace:

 Green curve: Input seismic

Red curve: Single-iteration RMS, with window length 21 samples. Note the stair-step appearance.

Blue curve: three iterations of RMS, with window length 11 samples. Note the...