Low-rank diffraction separation using an improved MSSA algorithm.

Diffraction separation is a vital key step for high-resolution imaging of small-scale geologic discontinuities. The multichannel singular-spectrum analysis (MSSA) algorithm has exhibited adequacy for diffraction separation and reflection suppression. However, the traditional MSSA algorithm is expensive because of the singular-value decomposition (SVD) operator. In addition, the common rank misidentification renders the diffraction separation highly rank-sensitive. Random noise caused by energy leakage is another problem for diffraction separation that is unresolved by the conventional MSSA. In this study, we propose a prestack diffraction separation method involving an improved MSSA algorithm. The new algorithm enables faster singular value estimation relative to the conventional SVD, with a diagonal Hankelization operator for artificial linear and random noise suppression, thereby enhancing the signal-to-noise ratio. Rank misidentification is alleviated by lowering the sensitivity to rank of the separated diffractions. Synthetic and field data are utilized to demonstrate the feasibility and superiority of the proposed method in computational efficiency and noise suppression compared with the conventional method. [ABSTRACT FROM AUTHOR]