Description and simulation of an active imaging technique utilizing two speckle fields: iterative reconstructors.

Abstract

Quasi-monochromatic light will form laser speckle upon reflection from a rough object. This laser speckle provides information about the shape of the illuminated object. In a prior paper [J. Opt. Soc. Am. A 19, 444 (2002)], it was shown that two intensities of two speckle patterns and their interference are sufficient to produce an unambiguous (except for object translation) band-limited image of the object, based on a root-matching technique described therein, in the absence of measurement error and in the case of distinct roots of the field polynomials and their complex conjugates. On the other hand, algorithms based on the root-matching technique are found to be slow and sensitive to noise. So motivated, several other techniques are applied to the problem, including phase retrieval, expectation maximization, and statistical maximization. The phase-retrieval and expectation-maximization techniques proved to be most effective for reconstructions of complex objects larger than 10 pixels across, and high-quality images were formed by using three independent sets of two-field data (three frames of two-wavelength data), each comprising two speckle intensity patterns and their interference. Two additional results of note are reported. First, the expectation-maximization algorithm produced relatively good images when three or more frames each of only one speckle intensity pattern (data at just one wavelength) were used and second, the phase-retrieval algorithm when only the object autocorrelation was used also produced relatively good images for the chosen test object.