Reconstruction of signals from magnitudes of redundant representations: the complex case (Q300888): Difference between revisions

Novel necessary conditions for complex-valued signal reconstruction from magnitudes of frame coefficients are presented. Deterministic stability bounds (Lipschitz constants) and stochastic performance bounds (Cramer-Rao lower bound) are presented. The entire analysis is done canonically that is independent of a particular choice of basis. Then an optimization algorithm based on the least-square error is proposed and analyzed. The algorithm performance is compared to the theoretical lower bound given by the Cramer-Rao inequality.