Multisegment Parallel Coregistration of Sentinel-1 SAR Time-Series Images by Combining OpenMP With MPI

Sentinel-1 synthetic aperture radar (SAR) imagery, with its wide-swath imaging capabilities, provides crucial data for large-area interferometric SAR (InSAR) surface deformation monitoring. Precise image coregistration is essential but computationally intensive, especially for time-series processing. Current methods leveraging OpenMP for algorithm-level parallelism struggle with the large swaths and data volumes of Sentinel-1, limiting efficiency for practical applications. To address these challenges, this study introduces a multisegment parallel coregistration method combining OpenMP and MPI. The geometric coregistration process uses OpenMP and MPI for algorithm-level and multitask parallelism, boosting efficiency. In the enhanced spectral diversity coregistration, slave images are segmented by temporal baseline, with the first image processed serially and others in parallel. Validation with 30 Sentinel-1 SAR images from Tianjin-Tangshan (plain) and Zhejiang (mountainous) regions demonstrates significant improvements. OpenMP achieves optimal efficiency with 10-12 parallel kernels, while the combined OpenMP-MPI method performs best with 16 tasks for geometric coregistration and 24 for enhanced spectral diversity coregistration. The proposed method improves processing speeds by 8.2 and 6.7 times in plain regions and achieves comparable gains in mountainous areas, surpassing OpenMP alone. This approach effectively meets the efficiency demands of Sentinel-1 time-series coregistration.