Airborne SAR-Efficient Signal Processing for Very High Resolution

Frequency-domain synthetic aperture radar (SAR) image formation algorithms are of lower computation cost (both in number of elementary operations and in required memory storage) than direct time-domain integration, and do not make the narrowband (monochromatic) assumption. Both advantages are critical to very-high-resolution imaging because a lower complexity yields a drastic computation time decrease as cross-range resolution increases, and the narrowband assumption is more and more a concern as range resolution (hence bandwidth) increases. Though an exact formulation exists (omega-k algorithm) for a perfect linear uniform acquisition trajectory, in a real-life airborne case, the unavoidable trajectory deviation from a straight line needs to be compensated. This motion compensation (MoComp) operation is much more complicated in the case of frequency-domain processing. An efficient technique for this purpose is presented. This method keeps the parallel processing aspect, and has been programmed both for multithread on multicore/ symmetrical multiprocessor central processing units (CPUs) and for graphic processor units (GPUs).