SPECTRAL WHITENING in practice

Contents

• Introduction
• Theory
• Types of Spectral Whitening

Introduction

Spectral Whitening (sometimes called balencing or broadening) is a process usually applied post-migration to improve the resolution and appearance of seismic data and is a crude attempt to correct for frequency attenuation. A variety of methods may be used to attempt to equalise the spectrum without overly boosting noise.

Theory

The theory of spectral whitening is described by the adjacent figure showing an input amplitude spectrum (a) of (typically) a migrated section. The objective of spectral whitening is to boost the balance of frequencies ultimately to obtain perfect resolution as shown by the green line in (b). While this spectrum could in theory be obtained, in practise it would likely result in the boosting of noise at low and high frequencies. More sensible would be to whiten the data within its own bandwidth as shown schematically by the blue curve.

return to contents

Types of Spectral Whitening

1. Inverse Filtering would tend to produce the green line in the previous figure. Some implementations may allow a percentage of the input data to be added back to stabilise the results.
2. The commonest form of spectral whitening is to split the dataset into several narrow frequency bands by bandpass filtering, equalising the sections by AGC (or some other scaling function) and add the resulting sections together. The user can chose the number of filter bands (i.e. select the output spectrum) and the degree of scaling applied. Some applications allow a percentage of the input data to be added back to improve the results.

return to contents