Synthetic aperture radar autofocus based on a bilinear model

Area of Science:

• Signal Processing
• Remote Sensing
• Image Restoration

Background:

• Nonideal synthetic aperture radar (SAR) imaging systems require autofocus algorithms for image restoration.
• Existing autofocus methods, like multichannel autofocus (MCA), have limitations in handling complex image models and wider look angle ranges.

Purpose of the Study:

• To propose a novel bilinear parametric model for image and phase parameter estimation.
• To derive an efficient maximum-likelihood autofocus (MLA) algorithm as a generalization of MCA.
• To analyze the advantages of MLA over previous MCA extensions regarding identifiability and noise sensitivity.

Main Methods:

• Development of a bilinear parametric model for SAR image and nuisance phase parameters.
• Derivation of a maximum-likelihood autofocus (MLA) algorithm.
• Numerical approximation of the constant modulus quadratic program central to these algorithms.
• Computer simulations to evaluate performance under various system models.

Main Results:

• The proposed MLA algorithm generalizes MCA to a broader class of models and look angle ranges.
• MLA demonstrates improved identifiability conditions and reduced noise sensitivity compared to prior extensions.
• Numerical approximations for the constant modulus quadratic program are proposed.
• Simulations confirm superior performance of MLA in terms of mean squared error and visual image quality.

Conclusions:

• The developed MLA algorithm offers a significant advancement in SAR image autofocus.
• MLA provides enhanced robustness and performance, particularly in complex imaging scenarios.
• The proposed numerical approximations facilitate the practical implementation of advanced autofocus techniques.