Topics in image-formation algorithms for synthetic aperture radar

In this dissertation, we explore topics in image-formation algorithms for synthetic aperture radar (SAR). SAR image formation requires spatially variant processing to generate a properly focused image. Various image-formation algorithms exist that approximate this spatially variant processing with a spatially invariant form for computational efficiency. In some SAR applications, this inaccurate approximation results in severe image artifacts.; Our first topic explores the performance of spatially variant apodization (SVA) for image reconstruction from partial Fourier data. We formulate SVA as a window for harmonic analysis and establish a connection with minimum-variance spectral estimation (MVSE). We study the performance of various 2-D SVA techniques through controlled simulations and well-defined performance measures.; We next consider the effect of a nonplanar wavefront in the tomographic formulation of spotlight-mode SAR. The tomographic formulation is based on a plane-wave assumption. This requires the dimension of the target scene to be much smaller than the range to the target. We analyze this problem for highly squinted SAR, by incorporating higher-order terms in the phase approximation.; The wavenumber-domain algorithm is based on an accurate approximation of wave-front curvature. A straightforward derivation of the {dollar}omega-k{dollar} algorithm is developed that is based on a quadratic phase approximation. We present an alternative wavenumber-domain algorithm for broadside SAR that is based on a range approximation customarily used in SAR.; Although the {dollar}omega-k{dollar} algorithm works well in most SAR scenarios, the reconstructed image suffers from artifacts as the radar squint angle approaches 90{dollar}spcirc.{dollar} We propose a novel, general-inversion algorithm for strip-mapping SAR. Unlike conventional SAR image-formation algorithms, our approach is more general without explicit approximations of the range function, and is robust to measurement noise.; Finally, we investigate image formation for highly squinted SAR that arises in runway imaging from an approaching aircraft. We explore theoretical feasibility of the runway-imaging scenario. We apply two image-formation algorithms: our general-inversion algorithm and the {dollar}omega-k{dollar} algorithm. The general-inversion algorithm achieves resolution consistent with the range of viewing angles.