Advertisement

What Do Features Tell about Images?

  • Mads Nielsen
  • Martin Lillholm
Conference paper
Part of the Lecture Notes in Computer Science 2106 book series (LNCS, volume 2106)

Abstract

According to the Marr paradigm [10], visual processing is performed by low-level feature detection followed by higher level task dependent processing. In this case, any two images exhibiting identical features will yield the same result of the visual processing. The set of images exhibiting identical features form an equivalence class: a metameric class [7]. We choose from this class the (in some precise sense) simplest image as a representative. The complexity of this simplest image may in turn be used for analyzing the information content of features. We show examples of images reconstructed from various scale-space features, and show that a low number of simple differential features carries sufficient information for reconstructing images close to identical to the human observer. The paper presents direct methods for reconstruction of minimal variance representatives, and variational methods for computation of maximum entropy and maximum a posteriori representatives based on priors for natural images. Finally, conclusions on the information content in blobs and edges are indicated.

Keywords

Feature Strength Natural Image Edge Point Image Code Simple Image 
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.
    J.H. Elder and S.W. Zucker. Scale space localization, blur, and contour-based image coding. In CVPR96, pages 27–34, 1996.Google Scholar
  2. 2.
    D.J. Field. Relations between the statistics of natural images and the response proporties of cortical cells. J. Optic. Soc. Am., 4(12):2379–2394, Dec 1987.CrossRefGoogle Scholar
  3. 3.
    J. Huang, A.B. Lee, and D. Mumford. Statistics of range images. In CVPR00, pages I:324–331, 2000.Google Scholar
  4. 4.
    R.A. Hummel and R. Moniot. Reconstructions from zero-crossings in scale-space. ASSP, 37(12):2111–2130, December 1989.CrossRefGoogle Scholar
  5. 5.
    Peter Johansen, Mads Nielsen, and Ole F Olsen. Branch points in one-dimensional gaussian scale-space. JMIV, 2000. To appear.Google Scholar
  6. 6.
    J.J. Koenderink. The structure of images. BioCyber, 50:363–370, 1984.MathSciNetzbMATHGoogle Scholar
  7. 7.
    J.J. Koenderink and A.J. van Doorn. Metamerism in complete sets of image operators. In AIU96, pages 113–129, 1996.Google Scholar
  8. 8.
    T. Lindeberg. On scale selection for differential operators. In ISRN KTH, 1993.Google Scholar
  9. 9.
    T. Lindeberg. Feature detection with automatic scale selection. IJCV, 30(2):79–116, November 1998.CrossRefGoogle Scholar
  10. 10.
    D. Marr. Vision: A computational investigation into the human representation and processing of visual information. In W.H. Freeman, 1982.Google Scholar
  11. 11.
    K.S. Pedersen and M. Nielsen. The hausdorff dimension and scale-space normalization of natural images. JVCIR, 11(2):266–277, June 2000.Google Scholar
  12. 12.
    D.L. Ruderman and W. Bialek. Statistics of natural images: Scaling in the woods. Physical Review Letters, 73(6):100–105, August 1994.CrossRefGoogle Scholar
  13. 13.
    S.J. Wernecke and L.R. d’Addario. Maximum entropy image reconstruction. TC, 26(4):351–364, April 1977.zbMATHGoogle Scholar
  14. 14.
    A.P. Witkin. Scale space filtering: A new approach to multi-scale descriptions. In IU84, pages 79–95, 1984.Google Scholar
  15. 15.
    Y.N. Wu, S.C. Zhu, and X.W. Liu. Equivalence of julesz and gibbs texture ensembles. In ICCV99, pages 1025–1032, 2000.Google Scholar
  16. 16.
    S.C. Zhu, Y. Wu, and D. Mumford. Minimax entorpy principles and its application to texture modeling. NeurComp, 9(8):1627–1660, 1997.Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2001

Authors and Affiliations

  • Mads Nielsen
    • 1
  • Martin Lillholm
    • 1
  1. 1.IT University of CopenhagenCopenhagenDenmark

Personalised recommendations