Nonuniform Grid Fast Factorized Backprojection for Multimode SAR Processing

Abstract

Conventional fast factorized backprojection (FFBP) algorithms for synthetic aperture radar (SAR) modes such as stripmap, sliding spotlight, and TOPS suffer from oversampling redundancy in partially illuminated regions due to the use of uniform azimuth grids, which limits computational efficiency. This article proposes a nonuniform grid FFBP (NUGFFBP) algorithm to reduce this redundancy. First, a unified azimuth wavenumber spectrum bandwidth variation model is established for the three modes, and the mechanism of oversampling redundancy is quantitatively analyzed. Second, a nonlinear grid stretching transform is designed to generate a nonuniform imaging grid that adaptively matches the local wavenumber spectrum bandwidth while retaining a configurable protection margin. Third, to address the high sensitivity of nonuniform grids to spectral misalignment, a modified spectral alignment method embedded in the backprojection (BP) kernel is proposed to ensure high-precision coherent fusion of nonuniform grid images. Finally, the effectiveness of the algorithm is verified through point-target simulations and real SAR data. Experimental results show that, while maintaining imaging quality comparable to conventional FFBP, NUGFFBP improves overall imaging efficiency by 7%–15% for the test cases, demonstrating excellent multimode adaptability.