A three-parameter high-resolution 3D fast Radon Transform

Abstract

When the underground medium exhibits anisotropy, the time slices of seismic data after normal moveout correction exhibit non-standard elliptical shapes. This complexity hinders the accurate representation by the traditional two-parameter 3D Radon transforms. To address this limitation, we introduced an additional parameter to describe ellipse rotation and integrated the frequency and curvature together, resulting in a frequency-independent transform operator, which can enhance both accuracy and computational efficiency of the algorithm. However, the algorithm suffers from low resolution in the transform domain, which limits its effectiveness in suppressing multiples with small moveouts. Existing high-resolution fast Radon transform algorithms typically adopt an iterative approach to obtain respective regularization terms for each frequency, which improves resolution but at the cost of sacrificing the frequency-independent advantage of the fast algorithm. This significantly increases computational costs. In response, we propose a novel three-parameter high-resolution 3D fast Radon transform algorithm based on the constraints derived from dominant frequency calculation. This algorithm utilizes calculation results from dominant frequency seismic data as the regularization term for all frequency calculations. By pre-computing high-resolution transform operator and its inversion, and calling them in all frequency calculations, the algorithm can effectively enhance the suppression of multiples with small moveouts while avoiding a significant increase in computational costs. To further enhance computational efficiency, we also implement a GPU-based strategy for the algorithm, which significantly improves the practicality of the approach. Both synthetic and real data tests have verified the feasibility and effectiveness of this new method.