Transformation-Invariant Embedding for Image Analysis

  • Ali Ghodsi
  • Jiayuan Huang
  • Dale Schuurmans
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3024)


Dimensionality reduction is an essential aspect of visual processing. Traditionally, linear dimensionality reduction techniques such as principle components analysis have been used to find low dimensional linear subspaces in visual data. However, sub-manifolds in natural data are rarely linear, and consequently many recent techniques have been developed for discovering non-linear manifolds. Prominent among these are Local Linear Embedding and Isomap. Unfortunately, such techniques currently use a naive appearance model that judges image similarity based solely on Euclidean distance. In visual data, Euclidean distances rarely correspond to a meaningful perceptual difference between nearby images. In this paper, we attempt to improve the quality of manifold inference techniques for visual data by modeling local neighborhoods in terms of natural transformations between images—for example, by allowing image operations that extend simple differences and linear combinations. We introduce the idea of modeling local tangent spaces of the manifold in terms of these richer transformations. Given a local tangent space representation, we then embed data in a lower dimensional coordinate system while preserving reconstruction weights. This leads to improved manifold discovery in natural image sets.


Face Image Principle Component Analysis Target Image Visual Data Local Linear Embed 
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.


  1. 1.
    Tenenbaum, J., de Silva, V., Langford, J.: A global geometric framework for nonlinear dimensionality reduction. Science 290, 2319–2323 (2000)CrossRefGoogle Scholar
  2. 2.
    Saul, L., Roweis, S.: Think globally, fit locally: Unsupervised learning of nonlinear manifolds. JMLR (2003)Google Scholar
  3. 3.
    Cox, T., Cox, M.: Multidimensional Scaling, 2nd edn. Chapman Hall, Boca Raton (2001)zbMATHGoogle Scholar
  4. 4.
    Jolliffe, I.: Principal Component Analysis. Springer, Heidelberg (1986)Google Scholar
  5. 5.
    Lebanon, G.: Learning Riemannian metrics. In: Proceedings UAI (2003)Google Scholar
  6. 6.
    Belkin, M., Niyogi, P.: Laplacian eigenmaps and spectral techniques for embedding and clustering. In: Proceedings NIPS (2001)Google Scholar
  7. 7.
    Teh, Y.-W., Roweis, S.: Automatic alignment of local representations. In: Proceedings NIPS (2002)Google Scholar
  8. 8.
    Hinton, G., Roweis, S.: Stochastic neighbor embedding. In: Proc. NIPS (2002)Google Scholar
  9. 9.
    Frey, B., Jojic, N.: Estimating mixture models of images and inferring spatial transformations using the em algorithm. In: Proceedings CVPR (1999)Google Scholar
  10. 10.
    Frey, B., Jojic, N.: Transformed component analysis: joint estimation of spatial transformations and image components. In: Proceedings ICCV, vol. 2, pp. 1190–1196 (1999)Google Scholar
  11. 11.
    Frey, B., Jojic, N.: Transformation-invariant clustering and dimensionality reduction using em. To appear in IEEE PAMI (2003)Google Scholar
  12. 12.
    Simard, P., Le Cum, Y., Denker, J.: Efficient pattern recognition using a new transformation distance. In: Proceedings NIPS 5 (1993)Google Scholar
  13. 13.
    Black, M., Jepson, A.: EigenTracking: Robust matching and tracking of articulated objects using a view-based representation. International Journal of Computer Vision 26(1) (1998)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2004

Authors and Affiliations

  • Ali Ghodsi
    • 1
  • Jiayuan Huang
    • 1
  • Dale Schuurmans
    • 2
  1. 1.School of Computer ScienceUniversity of Waterloo 
  2. 2.Department of Computing ScienceUniversity of Alberta 

Personalised recommendations