A Comparison of Autofocus Algorithms for SAR Imagery

A challenge in SAR system development involves compensation for nonlinear motion errors of the sensor platform. The uncompensated along-track motions can cause a severe loss of geometry accuracy and degrade SAR image quality. Autofocus techniques improve image focus by removing a large part of phase errors present after conventional motion compensation. It refers to the computer-automated error estimation and subsequent removal of the phase errors. Many autofocus algorithms have been proposed over the years, ranging from quantitative measurement of residual errors to qualitative visual comparison. However, due to the fact that different data sets and motion errors were employed, it is difficult to perform comparative studies on various algorithms. This paper compares and discusses some practical autofocus algorithms by using a common data set. Standard focal quality metrics are defined to measure how well an image is focused. Their implementation schemes and performance are evaluated in the presence of various phase errors, which include polynomial-like, high frequency sinusoidal, and random phase noise.