Parallel Processing of Satellite-Borne SAR Data for Accurate and Efficient Image Formation

Inhalt:
Image generation from satellite-borne synthetic aperture radar (SAR) video data requires numerical computations of the order of gigaflops (GFLOPs). Current parallel processing schemes distribute computation load among a number of parallel processing elements (PEs) for improving efficiency in computation. In this paper, we report a parallel signal processing algorithm for achieving very high efficiency in numerical computations of SAR image formation. We also demonstrate a two-dimensional memory mapping technique for range-cell migration correction (RCMC) that reduces substantially the number of memory call operations for re-indexing Doppler data. Sample SAR images from RADARSAT-1 video data are generated to validate accuracy and efficiency in image formation by parallel processing.