Random noise cancellation in seismic data using a 3D adaptive median filter

Abstract

Summary Median filters have been successfully applied to seismic data to reduce random noise. Due to their edge-preserving nature, median filters are also sometimes applied prior to edge detection. Current median filters use a fixed size which does not account for either the amount of local noise or local structure. To enhance the edge-preserving capability of median filters we incorporate a noise estimation criterion in the filter design and then adaptively change the 3D filter size based on local noise levels. This adaptive median filter locally balances the removal of random noise with the preservation of important geological features such as faults, fractures, and channels. The adaptive median filter is compared with a classic fixed-size median filter as a preprocessor to edge detection using the coherence cube algorithm. The adaptive median filter shows improved signal-to-noise ratio and better edge preservation.