Phase Linking (PL) techniques face challenges in constructing high-quality Sample Covariance Matrices (SCM) for Distributed Scatterers (DS), due to limited Statistically Homogeneous Pixels (SHPs) and possible heterogeneity. To address this, this paper proposes a method termed SE-NS, which integrates Spectral Entropy (SE) weighting and Nonlinear Shrinkage (NS). SE-NS assigns weights based on spectral entropy differences between a pixel and its SHPs, applies weighted averaging to the SCM, and then performs nonlinear shrinkage on the Sample Coherence Magnitude Matrix (SCMM) to obtain an optimized SCM and SCMM. The EMI algorithm is subsequently used for DS time-series phase estimation. Tests on simulated and real SAR data show that SE-NS outperforms traditional PL in noise suppression and phase smoothing, improves phase quality for deformation retrieval, and increases effective DS points by ~7% compared to EMI.
Yu et al. (Thu,) studied this question.