Nonlocal Phase Linking for Distributed Scatterer Interferometry

Distributed scatterer (DS) is prone to be seriously disturbed in differential interferometric process due to spatiotemporal decorrelation noise. Phase linking (PL) plays a pivotal role in DS interferometry (DSI) as it facilitates the recovery of the DS phase series from the temporal interferogram stack. Currently, the existing PL algorithms mainly rely on the statistically homogeneous pixels (SHPs) within a fixed patch to construct the sample complex coherence matrix (SCM). Nevertheless, these methods suffer from issues, such as the absence of spatial constraints, insufficient utilization of possible SHPs outside the window, and vulnerability to heterogeneous samples. To overcome the aforementioned limitations, this article proposes a nonlocal phase linking (NL-PL) approach inspired by the concept of NL means. NL-PL converts the resemblance between individual pixels and their homogenous neighboring counterparts into weights, which are subsequently used to perform a weighted average of the SCM. The weighted SCM is then employed in PL to achieve the optimization estimation of DS phase. The performance of the proposed method is evaluated using simulated data as well as real Sentinel-1 satellite data. The experimental results suggest that in contrast to conventional PL techniques, the proposed NL-PL approach showcases enhanced efficacy in suppressing noise and smoothing phase. Moreover, it enables a higher posterior coherence, thereby augmenting the count of DS sampling points. Furthermore, the findings of this study prove the performance and the adaptability of NL-PL in scenarios characterized by a scarcity of homogeneous samples and an abundance of heterogeneous samples.