Partitioned inverse image reconstruction for millimeter-wave SAR imaging

Synthetic aperture radar (SAR) images are representations of the microwave or millimeter-wave reflectivity of the observed scenes. SAR image reconstruction is an inverse problem, which can be solved via an approximation, e.g. matched filter (MF), or the explicit inverse using a large amount of measurement data. However, the approximation limits the resolution while the explicit inverse is computationally complex and mostly ill-conditioned. This paper proposes a partitioned inverse (PI) approach based on the Moore-Penrose pseudo inverse using truncated singular value decomposition for regularization, which is robust to noise. It is shown that PI has an improved resolution of 24% over MF even at 0 dB SNR and is three orders of magnitude faster than the explicit inverse. A measurement based proof of concept experiment using a laboratory K-Band (15–26.5 GHz) ultra-wideband SAR system is shown to validate the proposed approach.