Seismic random noise reduction by empirical mode decomposition combined with translation invariant scale-adaptive threshold

This paper applies empirical mode decomposition (EMD) combined with translation invariant (TI) scale-adaptive soft-hard compromise threshold method to achieve seismic random noise reduction. Commonly, the Gibbs phenomenon produced by wavelet transform is derived from the basis function which leads to artifacts in discontinuities. TI is a method which could suppress Gibbs artifacts efficiently and it is usually used in wavelet threshold denoising. We adopt threshold method in EMD for the reason that there is modes-mixing phenomenon in the decomposition procedure and if we discard the high frequency components directly, some valid signal components may be lost. Although EMD requires no basis function, TI in EMD threshold denoising method could get effective results especially aiming at block and heavy noise in seismic records. We take synthetic seismic signal with Gaussian white noise as experiment to test the good denosing ability of the new method and process real seismic records with block and heavy noise to prove the effectiveness of TI.