Abstract

We report on a new method for image recovery from speckle patterns, based on a complex phase retrieval method. Earlier work discussed both the theory¹ and experimental demonstration of image recovery² from speckle patterns. The previously reported work processed each speckle pattern to estimate autocorrelation, averaged autocorrelation estimates to reduce noise, and then did phase retrieval from the power spectral density to recover the image. The current work uses phase retrieval to directly recover a complex image field from each speckle pattern. The magnitudes of each recovered image are then averaged. The resulting imagery shows improved signal-to-noise ratio (SNR) compared with previous work. The use of sufficiently tight support constraints allows reconstruction of imagery with an SNR equal to the square root of the number of the speckle patterns processed. Improved SNR means that the required laser illumination energy may be reduced in long-range high resolution imaging systems based on speckle pattern sampling.

© 1989 Optical Society of America