The Phase Gradient Autofocus (PGA) algorithm can remove the high order phase error
effectively, which is of great significance to get high resolution images in real-time
processing. While PGA usually needs iteration, which necessitates long working hours.
In addition, the performances of the algorithm are not stable in different scene applications.
This severely constrains the application of PGA in real-time processing. Isolated
scatter selection and windowing are two important algorithmic steps of Phase Gradient
Autofocus Algorithm. Therefore, this paper presents an isolated scatter selection
method based on sample mean and a windowing method based on pulse envelope. These
two methods are highly adaptable to data, which would make the algorithm obtain better
stability and need less iteration. The adaptability of the improved PGA is demonstrated
with the experimental results of real radar data.