Abstract
A method of image recovery using noniterative phase retrieval is proposed and investigated by simulation. This method adapts the Cauchy–Riemann equations to evaluate derivatives of phase based on derivatives of magnitude. The noise sensitivity of the approach is reduced by employing a least-mean-squares fit. This method uses the analytic properties of the Fourier transform of an object, the magnitude of which is measured with an intensity interferometer. The solution exhibits the degree of nonuniqueness expected from root-flipping arguments for the one-dimensional case, but a simple assumption that restricts translational ambiguity also restricts the space of solutions and permits essentially perfect reconstructions for a number of nonsymmetric one-dimensional objects of interest. Very good reconstructions are obtained for a large fraction of random objects, within an overall image flip, which may be acceptable in many applications. Results for the retrieved phase and recovered images are presented for some one-dimensional objects and for different noise levels. Extensions to objects of two dimensions are discussed. Requirements for signal-to-noise ratio are derived for intensity interferometry with use of the proposed processing.
© 2004 Optical Society of America
Full Article | PDF ArticleMore Like This
R. B. Holmes, K. Hughes, P. Fairchild, B. Spivey, and A. Smith
J. Opt. Soc. Am. A 19(3) 444-457 (2002)
J. H. Seldin and J. R. Fienup
J. Opt. Soc. Am. A 7(3) 412-427 (1990)
Saowapak Sotthivirat and Jeffrey A. Fessler
J. Opt. Soc. Am. A 21(5) 737-750 (2004)