A New Median Filter for Suppressing Seismic Random Noise

Abstract

Random noise in seismic data is a major factor that affects signal-to-noise ratio, obscures details, and complicates identification of useful information. We present a new method for reducing random noise in seismic data. Called 1D time-varying median filter (TVMF), it is based on the 1D classical median filter (MF). We designed a threshold value that could control filter-window length according to characteristics of random noise. Using the relationship between seismic data and threshold value, we chose median filters with different time-varying filter-window lengths to eliminate random noise. We found that through processing the model and comparing it against two median filter methods, the time-varying median filter could strike a balance between eliminating random noise and protecting useful information. We used synthetic data to demonstrate the feasibility of our method in reducing seismic random noise. An example of synthetic data shows that TVMF is more effective in random noise suppression than the 2D multistage median filter (MLM) or MF. Results using the method on seismic data from the Songliao Basin in China demonstrate its effectiveness.