An iterative hybrid algorithm combining the steepest descent algorithm of Levenberg-Marquardt and the simulated annealing algorithm is used for reconstructing the complex permittivity of a two-dimensional, high-contrast, large object. The scattered fields calculated in a cost function are expanded in angular spectral modes. The ''illposedness'' in the inversion due to the noise in the scattered fields is regularized by keeping only the effective modes and filtering out the higher angular spectral modes. The object is discretized into N square cells of unknown permittivities and N or larger number of effective modes (P) are needed for this spectral domain iterative method to reconstruct the shape and the permittivity distribution of the object. When N>P, multiple view (K) and effective modes of each view (P) make the reconstruction possible if N≤PK. ¿ American Geophysical Union 1996