Advertisement

Abstract

Statistical analysis of spatially uniform signal contexts allows Discrete Universal Denoiser (DUDE) to effectively correct signal errors caused by a discrete symmetric memoryless transmission channel. The analysis sets no limits on a probability signal model apart from stationarity and ergodicity. Statistics of signal contexts are used first to learn the probability of errors and then to detect and correct the errors. Therefore a proper choice of context is an essential prerequisite to the practical use of DUDE. We propose to use the maximum likelihood estimate of context assuming the signals are modelled with a nonparametric generic Markov–Gibbs random chain or field. The model adds to stationarity and ergodicity only one more condition, namely, pairwise dependences between each signal and its context. Experiments with noisy binary images confirm a feasibility of such adaptive context, show some advantages of DUDE over more conventional median filtering, and relate the choice of a proper context size to the maximum entropy of the context statistics used for image denoising.

Keywords

Maximum Entropy White Area Image Denoising Signal Array Signal Context 
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.

References

  1. 1.
    Berman, M.: Automated smoothing of image and other regularly spaced data. IEEE Trans. Pattern Anal. Machine Intell. 16, 460–648 (1994)CrossRefGoogle Scholar
  2. 2.
    Blake, A., Zisserman, A.: Visual Reconstruction. MIT Press, Cambridge (1987)Google Scholar
  3. 3.
    Geman, S., Geman, D.: Stochastic relaxation, Gibbs distribution, and Bayesian restoration of images. IEEE Trans. Pattern Anal. Machine Intell. 6, 721–741 (1984)zbMATHCrossRefGoogle Scholar
  4. 4.
    Gimel’farb, G.L.: Image Textures and Gibbs Random Fields. Kluwer Academic, Dordrecht (1999)zbMATHGoogle Scholar
  5. 5.
    Rudin, L., Osher, S., Fatemi, E.: Nonlinear total variation based noise removal algorithm. Physica 60D, 259–268 (1992)Google Scholar
  6. 6.
    Weissman, T., Ordentlich, E., Seroussi, G., Verdu, S., Weinberger, M.: Universal Discrete Denoising: Known Channel. Technical Report HPL-2003-29. IT Research, HP Labs Palo Alto (February 10, 2003)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Georgy Gimel’farb
    • 1
  1. 1.Department of Computer Science, Tamaki CampusThe University of AucklandAucklandNew Zealand

Personalised recommendations