A Deep Boltzmann Machine-Based Approach for Robust Image Denoising

  • Rafael G. Pires
  • Daniel S. Santos
  • Gustavo B. Souza
  • Aparecido N. Marana
  • Alexandre L. M. Levada
  • João Paulo Papa
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 10657)


A Deep Boltzmann Machine (DBM) is composed of a stack of learners called Restricted Boltzmann Machines (RBMs), which correspond to a specific kind of stochastic energy-based networks. In this work, a DBM is applied to a robust image denoising by minimizing the contribution of some of its top nodes, called “noise nodes”, which often get excited when noise pixels are present in the given images. After training the DBM with noise and clean images, the detection and deactivation of the noise nodes allow reconstructing images with great quality, eliminating most of their noise. The results obtained from important public image datasets showed the validity of the proposed approach.


  1. 1.
    Papa, J.P., Fonseca, L.M.G., Carvalho, L.A.S.: Projections onto convex sets through particle swarm optimization and its application for remote sensing image restoration. Pattern Recognit. Lett. 31(13), 1876–1886 (2010)CrossRefGoogle Scholar
  2. 2.
    Pires, R.G., Pereira, D.R., Pereira, L.A.M., Mansano, A.F., Papa, J.P.: Projections onto convex sets parameter estimation through harmony search and its application for image restoration. Nat. Comput. 15(3), 493–502 (2016)MathSciNetCrossRefGoogle Scholar
  3. 3.
    Katsaggelos, A.K.: Digital Image Restoration. Springer, New York (1991)CrossRefGoogle Scholar
  4. 4.
    Paik, J.K., Katsaggelos, A.K.: Image restoration using a modified Hopfield network. IEEE Trans. Image Process. 1, 49–63 (1992)CrossRefGoogle Scholar
  5. 5.
    Sun, Y., Yu, S.Y.: A modified Hopfield neural network used in bilevel image restoration and reconstruction. In: International Symposium on Information Theory Application, vol. 3, pp. 1412–1414 (1992)Google Scholar
  6. 6.
    Keyvanrad, M.A., Pezeshki, M., Homayounpour, M.A.: Deep belief networks for image denoising. arXiv preprint arXiv:1312.6158 (2013)
  7. 7.
    Tang, Y., Salakhutdinov, R., Hinton, G.: Robust Boltzmann machines for recognition and denoising. In: IEEE Conference on Computer Vision and Pattern Recognition, Providence, Rhode Island, USA (2012)Google Scholar
  8. 8.
    Xie, J., Xu, L., Chen, E.: Image denoising and inpainting with deep neural networks. In: Pereira, F., Burges, C.J.C., Bottou, L., Weinberger, K.Q. (eds.) Advances in Neural Information Processing Systems, vol. 25, pp. 341–349. Curran Associates Inc. (2012)Google Scholar
  9. 9.
    Tang, Y., Salakhutdinov, R., Hinton, G.E.: Robust Boltzmann machines for recognition and denoising. In: IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012, pp. 2264–2271. IEEE (2012)Google Scholar
  10. 10.
    Yan, R., Shao, L.: Image blur classification and parameter identification using two-stage deep belief networks. In: 24th British Machine Vision Conference, pp. 1–11 (2013)Google Scholar
  11. 11.
    Osindero, S., Hinton, G.E., Teh, Y.W.: A fast learning algorithm for deep belief nets. Neural Comput. 18(7), 1527–1554 (2006)MathSciNetCrossRefMATHGoogle Scholar
  12. 12.
    Salakhutdinov, R., Hinton, G.E.: Deep Boltzmann machines. In: AISTATS, vol. 1, p. 3 (2009)Google Scholar
  13. 13.
    Hinton, G.E.: Training products of experts by minimizing contrastive divergence. Neural Comput. 14(8), 1771–1800 (2002)CrossRefMATHGoogle Scholar
  14. 14.
    Salakhutdinov, R., Hinton, G.E.: An efficient learning procedure for deep Boltzmann machines. Neural Comput. 24(8), 1967–2006 (2012)MathSciNetCrossRefMATHGoogle Scholar
  15. 15.
    Corinna, C., LeCun, Y., Burges, C.J.C.: The MNIST database of handwritten digits (1998)Google Scholar
  16. 16.
    Lichman, M.: UCI machine learning repository (2013)Google Scholar
  17. 17.
    Holub, A., Griffin, G., Perona, P.: Caltech-256 object category dataset (2007)Google Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2018

Authors and Affiliations

  1. 1.Department of ComputingUFSCar - Federal University of São CarlosSão CarlosBrazil
  2. 2.Department of ComputingUNESP - Univ Estadual PaulistaBauruBrazil

Personalised recommendations