Signal processing for microwave imaging systems with very sparse array

This dissertation investigates image reconstruction algorithms for near-field, two dimensional (2D) synthetic aperture radar (SAR) using compressed sensing (CS) based methods. In conventional SAR imaging systems, acquiring higher-quality images requires longer measuring time and/or more elements in an antenna array. Millimeter wave imaging systems using evenly-spaced antenna arrays also have spatial resolution constraints due to the large size of the antennas. This dissertation applies the CS principle to a bistatic antenna array that consists of separate transmitter and receiver subarrays very sparsely and non-uniformly distributed on a 2D plane. One pair of transmitter and receiver elements is turned on at a time, and different pairs are turned on in series to achieve synthetic aperture and controlled random measurements. This dissertation contributes to CS-hardware co-design by proposing several signal-processing methods, including monostatic approximation, re-gridding, adaptive interpolation, CS-based reconstruction, and image denoising. The proposed algorithms enable the successful implementation of CS-SAR hardware cameras, improve the resolution and image quality, and reduce hardware cost and experiment time. This dissertation also describes and analyzes the results for each independent method. The algorithms proposed in this dissertation break the limitations of hardware configuration. By using 16 x 16 transmit and receive elements with an average space of 16 mm, the sparse-array camera achieves the image resolution of 2 mm. This is equivalent to six percent of the λ/4 evenly-spaced array. The reconstructed images achieve similar quality as the fully-sampled array with the structure similarity (SSIM) larger than 0.8 and peak signal-to-noise ratio (PSNR) greater than 25 --Abstract, page iv