Advertisement

Abstract

Probabilistic models of textures should be able to synthesize specific textural structures, prompting the use of filter-based Markov random fields (MRFs) with multi-modal potentials, or of advanced variants of restricted Boltzmann machines (RBMs). However, these complex models have practical problems, such as inefficient inference, or their large number of model parameters. We show how to train a Gaussian RBM with full-convolutional weight sharing for modeling repetitive textures. Since modeling the local mean intensities plays a key role for textures, we show that the covariance of the visible units needs to be sufficiently small – smaller than was previously known. We demonstrate state-of-the-art texture synthesis and inpainting performance with many fewer, but structured features being learned. Inspired by Gibbs sampling inference in the RBM and the small covariance of the visible units, we further propose an efficient, iterative deterministic texture inpainting method.

Keywords

Hide Unit Deterministic Algorithm Textural Structure Texture Synthesis Texture Model 
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.
    Efros, A.A., Freeman, W.T.: Image quilting for texture synthesis and transfer. In: SIGGRAPH 2001 (2001)Google Scholar
  2. 2.
    Efros, A.A., Leung, T.K.: Texture synthesis by non-parametric sampling. In: ICCV 1999, vol. 2, pp. 1033–1038 (1999)Google Scholar
  3. 3.
    Geman, D., Reynolds, G.: Constrained restoration and the recovery of discontinuities. IEEE T. Pattern Anal. Mach. Intell. 14(3), 367–383 (1992)CrossRefGoogle Scholar
  4. 4.
    Hao, T., Raiko, T., Ilin, A., Karhunen, J.: Gated Boltzmann Machine in Texture Modeling. In: Villa, A.E.P., Duch, W., Érdi, P., Masulli, F., Palm, G. (eds.) ICANN 2012, Part II. LNCS, vol. 7553, pp. 124–131. Springer, Heidelberg (2012)CrossRefGoogle Scholar
  5. 5.
    Heess, N., Williams, C.K.I., Hinton, G.E.: Learning generative texture models with extended Fields-of-Experts. In: BMVC 2009 (2009)Google Scholar
  6. 6.
    Hinton, G.: A practical guide to training restricted Boltzmann machines. Tech. Rep. UTML TR 2010–003, University of Toronto (2010)Google Scholar
  7. 7.
    Kivinen, J.J., Williams, C.K.I.: Multiple texture Boltzmann machines. In: AISTATS (2012)Google Scholar
  8. 8.
    Luo, H., Carrier, P.L., Courville, A., Bengio, Y.: Texture modeling with convolutional spike-and-slab RBMs and deep extensions. In: AISTATS (2013)Google Scholar
  9. 9.
    Norouzi, M., Ranjbar, M., Mori, G.: Stacks of convolutional restricted Boltzmann machines for shift-invariant feature learning. In: CVPR 2009 (2009)Google Scholar
  10. 10.
    Portilla, J., Strela, V., Wainwright, M.J., Simoncelli, E.P.: Image denoising using scale mixtures of Gaussians in the wavelet domain. IEEE T. Image Process. 12(11), 1338–1351 (2003)CrossRefzbMATHMathSciNetGoogle Scholar
  11. 11.
    Ranzato, M., Mnih, V., Hinton, G.E.: Generating more realistic images using gated MRF’s. In: NIPS 2010 (2010)Google Scholar
  12. 12.
    Schmidt, U., Gao, Q., Roth, S.: A generative perspective on MRFs in low-level vision. In: CVPR 2010 (2010)Google Scholar
  13. 13.
    Tieleman, T.: Training restricted Boltzmann machines using approximations to the likelihood gradient. In: ICML 2008 (2008)Google Scholar
  14. 14.
    Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: From error visibility to structural similarity. IEEE T. Image Process. 13(4), 600–612 (2004)CrossRefGoogle Scholar
  15. 15.
    Welling, M., Hinton, G.E., Osindero, S.: Learning sparse topographic representations with products of Student-t distributions. In: NIPS 2002, pp. 1359–1366 (2002)Google Scholar
  16. 16.
    Zhu, S.C., Wu, Y., Mumford, D.: Filters, random fields and maximum entropy (FRAME): Towards a unified theory for texture modeling. Int. J. Comput. Vision 27(2), 107–126 (1998)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Qi Gao
    • 1
  • Stefan Roth
    • 1
  1. 1.Department of Computer ScienceTU DarmstadtGermany

Personalised recommendations