Abstract
Conventional gradient structure tensor (GST) for seismic data discontinuity detection inherently suffers from numerical instability at zero-crossings and artifacts in dipping strata. To address these limitations, we propose a coherence method based on a multiplicative analytic directional structure tensor (ADST), incorporating a robust eigenvector regularization strategy. First, to establish a reliable structural basis, we estimate local orientations using instantaneous phase (IP) attributes, applying an adaptive smoothing framework with sign-consistency correction. This step effectively resolves polarity ambiguities and suppresses noise, producing high-precision structural steering vectors. Subsequently, unlike the conventional directional structure tensor (DST), we construct the analytic tensor via matrix multiplication of the complex directional derivatives derived from the analytic signal. Theoretical analysis demonstrates that this multiplicative formulation induces a squaring effect on the eigenvalues. This mechanism nonlinearly amplifies the dominant eigenstructure while suppressing secondary eigenvalues associated with noise, thereby significantly enhancing sensitivity to discontinuities. Applications to synthetic and field data confirm that the proposed method eliminates zero-crossing instabilities and background artifacts in dipping strata. Crucially, the suppression of these artifacts reveals subtle faults and complex channel systems that were previously obscured by severe noise and stratigraphic interference. The resulting attribute delineates channel boundaries and fault edges with enhanced contrast and sharper definition, yielding a cleaner and more interpretable seismic coherence image.
Paper Information：
Z. Wang, J. Song, Y. Su, Z. Zhao, J. Shi and Z. Zhang, Eigenvector-Rectified Analytic Directional Structure Tensor for Seismic Data Discontinuity Detection, in IEEE Transactions on Geoscience and Remote Sensing, vol. 64, pp. 5916714-5916714, 2026, Art no. 5916714, doi: 10.1109/TGRS.2026.3720202. https://ieeexplore.ieee.org/document/11643272
