Generalized iterative sparse asymptotic minimum variance-based approach for SAR imaging

This paper proposes a generalized iterative Sparse Asymptotic Minimum Variance (GSAMV) framework for high-resolution imaging, demonstrated in the context of Synthetic Aperture Radar (SAR) reconstruction. The proposed method extends and unifies the existing SAMV family (SAMV-0, SAMV-1, and SAMV-2), effectively overcoming their individual limitations. While SAMV-0 produces highly sparse reconstructions and eliminates smearing, it degrades significantly under low-SNR conditions. SAMV-1 improves detection capability but suffers from reduced resolution, smearing, and outliers, whereas SAMV-2 provides a more balanced trade-off yet still struggles to simultaneously enforce sparsity while suppressing outliers and smearing. In contrast, the proposed adaptive GSAMV framework achieves an adaptive balance between sparsity, resolution, and the suppression of outliers and smearing through a tunable regularization parameter, enabling reliable reconstructions even under challenging low-SNR conditions. Furthermore, we propose an adaptive technique for selecting the optimal regularization parameter γ based on log-likelihood evaluation, which allows GSAMV to automatically adjust to different scenarios without manual tuning. The effectiveness of GSAMV is validated using both simulated and measured SAR data, demonstrating superior imaging performance compared to existing methods.