Auto-Learning of Parameters for High Resolution SAR Imagery

‍ ‍Aiming at the difficult problem of auto-learning of regularization term coefficients in high-resolution synthetic aperture radar （SAR） imaging based on regularization optimization， a marginal estimation Bayes （MEB） algorithm is proposed， so that the prior features of the target can be fitted properly to improve the accuracy of image feature extraction. Firstly， the alternating direction method of multipliers （ADMM） convex optimization framework is modeled based on the echoed data， and then the maximum marginal likelihood distribution of the parameters is derived. Moreover， the Moreau Yoshida unadjusted Langevin algorithm （MYULA） is used to realize the target extraction and the gradient projection method is introduced to estimate the regularization parameters. Finally， auto-learning parameters are used to optimize the imaging. The proposed algorithm can estimate the parameters of multiple regularization terms. Aiming at non-differentiable part in the prior， the subgradient optimization in the proximal method is used to solve the problem that gradient cannot solve through the adjacent operator， which can realize the parameter auto-learning of the non-differentiable regularization function. In the experimental part， compared with the optimal value of manual debugging， the error between the proposed method and the optimal value is within 15%. The effectiveness of the proposed algorithm is verified by phase transition diagram （PTD）. At the same time， this algorithm is compared with other auto-learning algorithms to verify the practicability of the algorithm.