Postprocessing tools for ultrawideband SAR images

This thesis provides an investigation of ultra-wideband wide angle (UWBWA) synthetic aperture radar (SAR) from a signal processing point of view. The approach taken is based on a system theoretic analysis of the radar imaging system, which consists of the physical data collection followed by image formation. The mathematical analysis is as general as possible with a minimum of approximations and simplifications. An overall analytic system model is obtained that reveals how each subsystem and each parameter of the imaging system contributes to formed SAR images. A thorough assessment of the system model proves that it describes actual UWBWA SAR systems closely.; The system model serves as a basis for further investigations of UWBWA SAR in several directions. First, it is shown that the particular shape of the frequency support of the collected UWBWA SAR data as well as of the formed images can be exploited in non-separable two-dimensional sampling schemes for data efficient implementations of UWBWA SAR algorithms. Properties of the system model indicate that in the case of UWBWA SAR image formation cannot be viewed anymore as the inversion of a linear system. Instead new approaches based on sophisticated data models are needed for analyzing UWBWA SAR data in a meaningful way. Furthermore, the system model uncovers some misconceptions found in the literature about the information content of images formed from UWBWA SAR data.; Many algorithms have been proposed in the literature that process the raw data prior to image formation to facilitate image interpretation or target detection and identification at later stages. A central result of this thesis is that many of these preprocessing algorithms can be translated into simple filtering operations of UWBWA SAR images. In particular filters that perform operations equivalent to aperture weighting, subaperture image formation, subbanding and resonance signature extraction are presented. All of these filters have a common structure and a filter design methodology tailored to this particular class of filters is introduced. It is a simple, yet flexible, non-iterative technique that leads to good quality filters. The performance of these filters is demonstrated on synthetic and real data.