Error-Diffusion Binarization for Neural Networks

  • André Granger
  • Tigran Galstyan
  • Roger A. Lessard

Abstract

In optical implementation of neural networks, binarization is often a necessity. Error-diffusion (ED) is presented has a more reliable binarization technic over hardclipping. We use a simple self-organizing learning algorithm for its demonstration. It is showed in particular that ED keeps more of the original image characteristics upon variable lighting, thus enabling a better extraction of the common features in a set of processed images, and a faster learning convergence of the neural net.

Keywords

Diffraction Efficiency Variable Lighting Gray Level Image Binarize Version Optical Implementation 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Kelvin Wagner and Demetri Psaltis. ‘Multilayer Optical Learning Networks’ Appl. Opt., 26, 23 5061–5075 (1987).CrossRefGoogle Scholar
  2. 2.
    Kelvin Wagner and Timothy M. Slagle. ‘Optical Competitive Learning With VLSI/Liquid Crystal Winner Take-All modulators’. Appl. Opt., 32, 8 (1993) 1408–1435.CrossRefGoogle Scholar
  3. 3.
    C.W. Slinger. ‘Weighted Volume Interconnections for Adaptive Networks’s Optical Computing and Processing’. 1, 3 (1991) 219–232.Google Scholar
  4. 4.
    Ken Y. Hsu, Pochi Yeh et al. ‘Conditionnal Convergence of Photorefractive Perceptron Learning’. Opt. Lett. 18, 24, (1993) 2135.CrossRefGoogle Scholar
  5. 5.
    P.J. Van Heerden, Appl. Opt. 2 (1963) 393.CrossRefGoogle Scholar
  6. 6.
    T. Galstyan, G. Pauliat, A. Villing and G. Roosen, ‘Adaptive Photorefractive Neurons for Self-Organizing Networks’, Optics Comm. 109 (1994) 35–42.CrossRefGoogle Scholar
  7. 7.
    D. Roberge, L. Gonçalves Neto and Y. Sheng ‘Programmable Optical Kinoforms and Complex-Amplitude Holograms with Coupled-Mode Modulation LCTV’ OE Reports July 1995.Google Scholar
  8. 8.
    R.W. Floyd and L. Steinberg, ‘An Adaptive Algorithm for Spatial Grayscale’, Proc. Soc. Inf. Disp. 17 (1976) 75–77.Google Scholar
  9. 9.
    R. Hauck and O. Bryngdahl. ‘Computer-Generated Holograms with Pulse-Density Modulation’ J. Opt. Soc. Am. A. 1 (1984) 5–10.CrossRefGoogle Scholar
  10. 10.
    E. Barnard ‘Optimal Error-Diffusion for Computer-Generated Holograms’ J. Opt. Soc. Am. A. 5 (1988) 1803–1817.CrossRefGoogle Scholar
  11. 11.
    R. EschBach and K. T. Knox, ‘Error-Diffusion Algorithm with Edge Enhancement’ J. Opt. Soc. Am. A. 8 (1991) 1844–1850.CrossRefGoogle Scholar
  12. 12.
    T. Kohonen ‘Self-Organization and Associative Memory’, 3rd Ed. Springer, Berlin, (1989).CrossRefGoogle Scholar
  13. 13.
    S. Grossberg, ‘Neural Networks and Natural Intelligence’, A Bradford Book, The Massachussets Institute of technology (1988).Google Scholar

Copyright information

© Springer Science+Business Media New York 1997

Authors and Affiliations

  • André Granger
    • 1
  • Tigran Galstyan
    • 1
  • Roger A. Lessard
    • 1
  1. 1.COPL, dépt. de PhysiqueUniversité LavalCanada

Personalised recommendations