May 2005
Volume 46, Issue 13
ARVO Annual Meeting Abstract  |   May 2005
Wavefront Reconstruction With Artificial Neural Network
Author Affiliations & Notes
  • J.F. Bille
    Physics, University of Heidelberg, Heidelberg, Germany
  • G. Hong
    Physics, University of Heidelberg, Heidelberg, Germany
  • N.A. Korablinova
    Physics, University of Heidelberg, Heidelberg, Germany
  • Footnotes
    Commercial Relationships  J.F. Bille, None; G. Hong, None; N.A. Korablinova, None.
  • Footnotes
    Support  None.
Investigative Ophthalmology & Visual Science May 2005, Vol.46, 1995. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      J.F. Bille, G. Hong, N.A. Korablinova; Wavefront Reconstruction With Artificial Neural Network . Invest. Ophthalmol. Vis. Sci. 2005;46(13):1995.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

Abstract: : Purpose: Wavefront measurements are widely used for many applications in adaptive optics, ranging from ophthalmology and astronomy to laser beam control. Usually the wavefront is reconstructed from the displacements of the spots through the least–square fit or the Single Value Decomposition (SVD) method. A new approach for the wavefront reconstruction is a method using Artifical Neural Networks. The goal of this work was a fundamental analysis and systematical study of the different network structures. Methods: Two types of the network architecture were studied: a two–layer network and a multiple–layer network to find an effective network structure. The training of the networks is performed with adequate learning schemes. The wavefront reconstruction made by the least–square fit and the Neural Networks was compared with respect to the precision of the wavefront reconstruction especially in cases of noisy measurement data. Results: The written software routines were used for diverse test measurements on the phase–plates with known optical aberrations and human eyes. The study has shown that by a simple network consisting of one input and one output layer the wavefront can be reconstructed even under difficult measurement conditions. The evaluation method using the Neural Networks has proven to be more precise than the least–square method if applied on very noisy data. Conclusions: The Artifical Neural Networks are very powerful tools for the wavefront reconstruction. Compared to the commonly used least–square fit there are advantages if noisy measurement data have to be evaluated.

Keywords: refraction • optical properties • refractive surgery: optical quality 

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.