Log transformation with Gauss-Newton microwave image reconstruction reduces incidence of local minima convergence

Microwave tomographic imaging has long been plagued by problems of converging to unwanted solutions or "local minima". For a variety of real- world system implementations, these issues were partially overcome by the use of good starting guesses or a priori information to the iterative reconstruction process. Poor convergence was recognized in early 2D imaging efforts and still persists as the technology progresses to 3D. These problems become progressively more pronounced at higher operating frequencies and for electrically large targets. At the opposite extreme - low contrast, small scatterers - linear approximations such as the Born and Rytov methods worked well.