Amplitude-Preserving 3-D TV Regularization for Seismic Random Noise Attenuation

Abstract

Conventional total variation (TV) regularization denoising model is typically constructed by the first-order differences in both lateral and vertical directions. However, first-order differences will result in poor amplitude-preserving outcomes for 3-D seismic random noise attenuation. To address this issue, in this letter, we reform the lateral- and vertical-related constraints in conventional TV regularization function based on high-order differences and Lagrange interpolation to adapt to the lateral and vertical features of seismic data, respectively. Then, we obtain our amplitude-preserving 3-D TV regularization method. In order to optimize the corresponding 3-D denoising objective function, we transform it into frequency-domain and propose a fast optimization method based on the split Bregman algorithm. Both synthetic and field data examples show that our proposed method can yield higher fidelity denoising results compared to the conventional approach.