The Shack-Hartmann sensor is used for wavefront sensing in adaptive optics. In the conventional use of the Shack-Hartmann sensor, information on the components other than wavefront tilt cannot be obtained. This means that the value of the Strehl ratio of the compensated image is not so high in spite of the fact that the Shack-Hartmann sensor images have higher order information about the global wavefront. We propose to use both the Shack-Hartmann sensor images and the image in the focal plane of the global wavefront as the constraints in the iterative Fourier transform algorithm proposed by J.R.Fienup. The reconstructed wavefront by using only the information on the gravity center of an image of the Shack-Hartmann sensor is used as an initial estimate in this iterative procedure. The effectiveness of this proposed algorithm is shown by computer simulation results. This proposed algorithm converged to an adequate wavefront rapidly without suffering from the stagnation problem and it showed good performance even in the case in which measured images were contaminated by photon noise.
The conventional iterative Fourier transform algorithm proposed by Fienup which reconstructs the phase from the modulus is improved. Stagnation from which the conventional algorithm often suffers does not occur in the improved algorithm.
An iterative algorithm to reconstruct an image from its Fourier modulus is proposed. Although this algorithm is a modification of the conventional iterative Fourier transform algorithm proposed by Fienup, stagnation which the conventional algorithm often suffers from does not occur in this algorithm.
A new technique for global wavefront sensing is proposed. This technique is based on the iterative phase retrieval algorithm proposed by Fienup. Shack-Hartmann sensor images are used as an additional constraint in this algorithm. Computer simulation results verified the effectiveness of this method.
We propose a blind deconvolution technique based on iterative Fourier transform algorithms. It has the convergence property although it needs many iterations and a support constraint. The deconvolution is accomplished completely in the noise free case. We also investigate its performance for the contaminated case.