Modified phase gradient algorithm for ISAR image autofocus with large apertures

Autofocus plays a major role in synthetic aperture radar (SAR) and inverse SAR (ISAR) processing to obtain well-focused radar images from the pulse data acquired by imaging radars. The simplest autofocus algorithm tracks a prominent (or dominant) scatterer and takes its phase as reference. If there are multiple scatterers with comparable amplitudes or if a dominant scatterer is not visible in the whole coherent processing interval (CPI), this algorithm often fails. The phase gradient algorithm (PGA) overcomes this situation by using multiple scatterers. In its original form it assumes that phase errors cause streaks in cross-range direction in the reconstructed image by one-dimensional convolution. This assumption is, however, only valid for short apertures. In this paper a modification of the PGA is proposed which works even for large apertures.