Estimation of Initial Guess of Steepest Descent Method for Near Field Phase Retrieval

A good initial guess is critical for the convergence of the steepest descent method. This paper presents a new method to estimate the initial guess of the near field phase. Based on image theory, tangential near fields on two scanning planes are linked by a radiation matrix. The radiation matrix is then scaled by the near field amplitudes. It is shown that the phase of an eigenvector of the scaled radiation matrix is a good approximation to the near field phase. The phase of the eigenvector is used as an initial guess, and it is further improved by the steepest descent method. Simulation results show that accurate phase retrieval is achieved by the proposed initial guess.