A Joint Framework for Seismic Signal Denoising Using Total Generalized Variation and Shearlet Transform

Seismic exploration is a remote-sensing tool applied in great many projects for engineering and resource-exploration purposes. Random noise suppression one of the key steps seismic-signal processing, especially those with important details features. The threshold-shrinkage method based on Shearlet transform has been effectively denoising. However, usually introduces boundary effect, which influences imaging quality. denoising total generalized variation (TGV) easy to produce `oil painting' but it can suppress effect. This paper proposes TGV making full use their characteristics, recover both edges fine much better than existing regularization methods. First, we result as input obtain primary residual profile. Second, interactive iteration extract signals efficiently from profile perform effective stack continuously. During adaptive-weight factor combined estimating optimal result. Last, final estimated obtained when stopping criterion met or maximum number iterations reached. synthetic field results show that proposed random noise. In addition, also remove effect further improves signal-to-noise ratio (SNR).