next up previous print clean
Next: Improving on existing methods Up: Vlad: Enhanced random noise Previous: Vlad: Enhanced random noise

Introduction

Spatial prediction filtering attenuates random noise uncorrelated from trace to trace, while preserving linear, predictable events. The prediction is formulated as a least-squares problem in either the t-x or the f-x domain. The methods are casually known as `` f-x decon'' and ``t-x decon.'' Although established by common practice, the name ``decon'' is not appropriate in these cases, because it suggests a similarity with the much better-known deconvolution of the signal along the time axis. However, deconvolution removes the predictable information (wavelet + multiples) and keeps the unpredictable (the reflectivity function), while f-x and t-x decons keep the predictable along the space axis (linear events), and remove the unpredictable (random noise). A very good explanation of spatial prediction filtering, with many examples, is given on page 960 of Yilmaz (2001). F-x decon was introduced by Canales (1984). Noise suppression in the t-x domain was developed by Abma (1995), within the framework of which the SEPlib programs Fx2d and Txdec were developed. I will show how to achieve Enhanced Random Noise Attenuation (ERNA) by building upon these existing spatial prediction filtering methods.
next up previous print clean
Next: Improving on existing methods Up: Vlad: Enhanced random noise Previous: Vlad: Enhanced random noise
Stanford Exploration Project
7/8/2003