Abstract

This paper presents a strategy for combining the results of image classification and image segmentation. The visual features used for classification and segmentation may be different in general. Fusion is performed in a Maximum Likelihood framework using the Expectation Maximization algorithm. Preliminary results show that segmentation may effectively contribute to increase the quality of classification.

Keywords

Posterior Distribution Conditional Independence Hard Segmenter Expectation Maximization Algorithm Neural Computation 
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.
    Breiman, L.: Bias, variance and arcing classifiers. Tech. Report 460, Statistics Department, UC Berkeley (1996)Google Scholar
  2. 2.
    Domingos, P.: A unified Bias-Variance decomposition for zero-one and squared loss. In: AAAI/IAAI, pp. 564–569 (2000)Google Scholar
  3. 3.
    Domingos, P., Pazzani, M.: On the optimality of the simple Bayesian classifier under zero–one loss. Machine Learning 29(2/3), 103–130 (1997)MATHCrossRefGoogle Scholar
  4. 4.
    Friedman, J.H.: On bias, variance, 0/1–loss, and the curse-of-dimensionality. Data Mining and Knowledge Discovery 1, 55–77 (1997)CrossRefGoogle Scholar
  5. 5.
    Geman, S.L., Bienenstock, E., Doursat, R.: Neural networks and bias/variance dilemma. Neural Computation 4, 1–58 (1992)CrossRefGoogle Scholar
  6. 6.
    Heskes, T.: Bias/Variance decomposition for likelihood–based estimators. Neural Computation 10, 1425–1433 (1998)CrossRefGoogle Scholar
  7. 7.
    Howe, N.R., Huttenlocher, D.P.: Integrating Color, Texture, and Geometry for Image Retrieval. In: Proc. Computer Vision and Patter Recognition, pp. 239–247 (2000)Google Scholar
  8. 8.
    Kittler, J., Hatef, M., Duin, R., Matas, J.: On combining classifiers. IEEE Trans. PAMI 20(3) (March 1998)Google Scholar
  9. 9.
    Kittler, J., Hojjatoleslami, A.A.: A weighted combination of classifiers employing shared and distinct representations. In: Proc. CVPR, pp. 924–929 (1998)Google Scholar
  10. 10.
    Kong, E.B., Dietterich, T.G.: Error–correcting output coding corrects bias and variance. In: Proc. Intl. Conf. Machine Learning, pp. 313–321 (1995)Google Scholar
  11. 11.
    Langley, P.: Induction of selective Bayesian classifiers. In: Proc. of the 10th Conference on Uncertainty in Artificial Intelligence, Seattle, WA (1994)Google Scholar
  12. 12.
    Manduchi, R.: Bayesian Fusion of Color and Texture Segmentations. In: 7th IEEE International Conference on Computer Vision, Kerkyra, September 1999, pp. 956–962 (1999)Google Scholar
  13. 13.
    Martin, J.K., Hirschberg, D.S.: Small sample statistics for classification error rates I: Error rate measurements, Dept. of Inf. and Comp. Sci., UC Irvine, Tech. Report 96–21 (1996)Google Scholar
  14. 14.
    Raudys, S., Jain, A.K.: Small sample size effects in statistical pattern recognition: Recommendations for practitioners. IEEE Trans. PAMI 13(3), 252–264 (1991)Google Scholar
  15. 15.
    Raudys, S.: On dimensionality, sample size, and classification error of nonparametric linear classification aglgorithms. IEEE Trans. PAMI 19(6), 337–371 (1997)Google Scholar
  16. 16.
    Ripley, B.D.: Pattern Recognition and Neural Networks. Cambridge University Press, Cambridge (1996)MATHGoogle Scholar
  17. 17.
    Shi, J., Malik, J.: Normalized Cuts and Image Segmentation. In: Proc. Computer Vision and Patter Recognition, pp. 731–737 (1997)Google Scholar
  18. 18.
    Shi, X., Manduchi, R.: A Study on Bayes Feature Fusion for Image Classification. In: IEEE Workshop on Statistical Algorithms for COmputer Vision (2003)Google Scholar
  19. 19.
    Tibshirani, R.: Bias, variance and predicition error for classification rules, Tech. Report, Dept. pf Prev. Medicine and Biosatistics, Univ, of Toronto (1996)Google Scholar
  20. 20.
    Wolpert, D.: On bias plus variance. Neural Computation 9, 1211–1243 (1997)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Roberto Manduchi
    • 1
  1. 1.University of California, Santa CruzSanta CruzUSA

Personalised recommendations