next up previous [pdf]

Next: 1-D stationary median filter Up: Liu etc.: 1-D time-varying Previous: Introduction

Theoretical basis

In contrast to useful information, random, spike-like noise in seismic data is neither continuous nor correlative, and a 1-D stationary median filter, having a large filter-window length, can easily remove such noise. However, signal can be damaged by such a filter.

The 1-D TVMF is based on the 1-D stationary median filter. We propose to measure the local noise content of the data and to adjust the filter length adaptively. If a threshold value that judges random noise and estimates noise intensity can be chosen, a filter having a large filter-window length can eliminate random noise while processing useful information by using a small filter-window length. Such a filter can thus effectively eliminate random noise while maximizing preservation of a detailed structure of useful information.