A Versatile Algorithm for Autofocusing SAR Images

Abstract

Introduction.Random deviations of the antenna phase centre of a synthetic aperture radar (SAR) are a source of phase errors for the received signal. These phase errors frequently cause blurring of the radar image. The image quality can be improved using various autofocus algorithms. Such algorithms estimate phase errors via optimization of an objective function, which defines the radar image quality. The image entropy and sharpness are well known examples of objective functions. The objective function extremum can be found by fast optimization methods, whose realization is a challenging computing task.Aim.To synthesize a versatile and computationally simple autofocusing algorithm allowing any objective function to used without changing its structure significantly.Materials and methods.An algorithm based on substituting the selected objective function with a simpler surrogate objective function, whose extremum can be found by a direct method, is proposed. This method has been referred as the MM optimization in scientific literature. It is proposed to use a quadratic function as a surrogate objective function.Results.The synthesized algorithm is straightforward, not requiring recursive methods for finding the optimal solution. These advantages determine the enhanced speed and stability of the proposed algorithm. Adjusting the algorithm for the selected objective function requires minimal software changes. Compared to the algorithm using a linear surrogate objective function, the proposed algorithm provides a 1.5 times decrease in the standard deviation of the phase error estimate, with an approximately 10 % decrease in the number of iterations.Conclusion.The proposed autofocusing algorithm can be used in synthetic aperture radars to compensate the arising phase errors. The algorithm is based on the MM-optimization of the quadratic surrogate objective functions for radar images. The computer simulation results confirm the efficiency of the proposed algorithm even in case of large phase errors.