Probabilistic mixture-based image modelling

Michal Haindl; Vojtěch Havlíček; Jiří Grim

Kybernetika (2011)

  • Volume: 47, Issue: 3, page 482-500
  • ISSN: 0023-5954

Abstract

top
During the last decade we have introduced probabilistic mixture models into image modelling area, which present highly atypical and extremely demanding applications for these models. This difficulty arises from the necessity to model tens thousands correlated data simultaneously and to reliably learn such unusually complex mixture models. Presented paper surveys these novel generative colour image models based on multivariate discrete, Gaussian or Bernoulli mixtures, respectively and demonstrates their major advantages and drawbacks on texture modelling applications. Our mixture models are restricted to represent two-dimensional visual information. Thus a measured 3D multi-spectral texture is spectrally factorized and corresponding multivariate mixture models are further learned from single orthogonal mono-spectral components and used to synthesise and enlarge these mono-spectral factor components. Texture synthesis is based on easy computation of arbitrary conditional distributions from the model. Finally single synthesised mono-spectral texture planes are transformed into the required synthetic multi-spectral texture. Such models can easily serve not only for texture enlargement but also for segmentation, restoration, and retrieval or to model single factors in unusually complex seven dimensional Bidirectional Texture Function (BTF) space models. The strengths and weaknesses of the presented discrete, Gaussian or Bernoulli mixture based approaches are demonstrated on several colour texture examples.

How to cite

top

Haindl, Michal, Havlíček, Vojtěch, and Grim, Jiří. "Probabilistic mixture-based image modelling." Kybernetika 47.3 (2011): 482-500. <http://eudml.org/doc/197057>.

@article{Haindl2011,
abstract = {During the last decade we have introduced probabilistic mixture models into image modelling area, which present highly atypical and extremely demanding applications for these models. This difficulty arises from the necessity to model tens thousands correlated data simultaneously and to reliably learn such unusually complex mixture models. Presented paper surveys these novel generative colour image models based on multivariate discrete, Gaussian or Bernoulli mixtures, respectively and demonstrates their major advantages and drawbacks on texture modelling applications. Our mixture models are restricted to represent two-dimensional visual information. Thus a measured 3D multi-spectral texture is spectrally factorized and corresponding multivariate mixture models are further learned from single orthogonal mono-spectral components and used to synthesise and enlarge these mono-spectral factor components. Texture synthesis is based on easy computation of arbitrary conditional distributions from the model. Finally single synthesised mono-spectral texture planes are transformed into the required synthetic multi-spectral texture. Such models can easily serve not only for texture enlargement but also for segmentation, restoration, and retrieval or to model single factors in unusually complex seven dimensional Bidirectional Texture Function (BTF) space models. The strengths and weaknesses of the presented discrete, Gaussian or Bernoulli mixture based approaches are demonstrated on several colour texture examples.},
author = {Haindl, Michal, Havlíček, Vojtěch, Grim, Jiří},
journal = {Kybernetika},
keywords = {discrete distribution mixtures; Bernoulli mixture; Gaussian mixture; EM algorithm; multi-spectral texture modelling; BTF texture modelling; EM algorithm; Bernoulli mixture; Gaussian mixture; multi-spectral texture modelling; BTF texture modelling; discrete distribution mixtures},
language = {eng},
number = {3},
pages = {482-500},
publisher = {Institute of Information Theory and Automation AS CR},
title = {Probabilistic mixture-based image modelling},
url = {http://eudml.org/doc/197057},
volume = {47},
year = {2011},
}

TY - JOUR
AU - Haindl, Michal
AU - Havlíček, Vojtěch
AU - Grim, Jiří
TI - Probabilistic mixture-based image modelling
JO - Kybernetika
PY - 2011
PB - Institute of Information Theory and Automation AS CR
VL - 47
IS - 3
SP - 482
EP - 500
AB - During the last decade we have introduced probabilistic mixture models into image modelling area, which present highly atypical and extremely demanding applications for these models. This difficulty arises from the necessity to model tens thousands correlated data simultaneously and to reliably learn such unusually complex mixture models. Presented paper surveys these novel generative colour image models based on multivariate discrete, Gaussian or Bernoulli mixtures, respectively and demonstrates their major advantages and drawbacks on texture modelling applications. Our mixture models are restricted to represent two-dimensional visual information. Thus a measured 3D multi-spectral texture is spectrally factorized and corresponding multivariate mixture models are further learned from single orthogonal mono-spectral components and used to synthesise and enlarge these mono-spectral factor components. Texture synthesis is based on easy computation of arbitrary conditional distributions from the model. Finally single synthesised mono-spectral texture planes are transformed into the required synthetic multi-spectral texture. Such models can easily serve not only for texture enlargement but also for segmentation, restoration, and retrieval or to model single factors in unusually complex seven dimensional Bidirectional Texture Function (BTF) space models. The strengths and weaknesses of the presented discrete, Gaussian or Bernoulli mixture based approaches are demonstrated on several colour texture examples.
LA - eng
KW - discrete distribution mixtures; Bernoulli mixture; Gaussian mixture; EM algorithm; multi-spectral texture modelling; BTF texture modelling; EM algorithm; Bernoulli mixture; Gaussian mixture; multi-spectral texture modelling; BTF texture modelling; discrete distribution mixtures
UR - http://eudml.org/doc/197057
ER -

References

top
  1. Bennett, J., Khotanzad, A., 10.1109/34.667889, IEEE Trans. Pattern Analysis and Machine Intelligence 20 (1998), 3, 327–332. (1998) DOI10.1109/34.667889
  2. Dana, K. J., Nayar, S. K., Ginneken, B. van, Koenderink, J. J., Reflectance and texture of real-world surfaces, In: CVPR, IEEE Computer Society, 1997, pp. 151–157. (1997) 
  3. Bonet, J. S. De, Multiresolution sampling procedure for analysis and synthesis of textured images, In: ACM SIGGRAPH 97, ACM Press, 1997, pp. 361–368. (1997) 
  4. Dempster, A. P., Laird, N. M., Rubin, D. B., Maximum likelihood from incomplete data via the em algorithm, J. Roy. Statist. Soc. B 39 (1977), 1, 1–38. (1977) Zbl0364.62022MR0501537
  5. Efros, A. A., Freeman, W. T., Image quilting for texture synthesis and transfer, In: ACM SIGGRAPH 2001 (E. Fiume, ed.), ACM Press, 2001, pp. 341–346. (2001) 
  6. Efros, A. A., Leung, T. K., Texture synthesis by non-parametric sampling, In: Proc. Internat. Conf. on Computer Vision (2), Corfu 1999, pp. 1033–1038. (1999) 
  7. Filip, J., Haindl, M., 10.1109/TPAMI.2008.246, IEEE Trans. Pattern Analysis and Machine Intelligence 31(2009), 11, 1921–1940. (2009) DOI10.1109/TPAMI.2008.246
  8. Grim, J., On numerical evaluation of maximum likelihood estimates for finite mixtures of distributions, Kybernetika 18 (1982), 173–190. (1982) Zbl0489.62028MR0680154
  9. Grim, J., Haindl, M., 10.1016/S0167-9473(02)00174-3, Comput. Statist. Data Anal. 43 (2003), 3–4, 603–615. (2003) MR1968071DOI10.1016/S0167-9473(02)00174-3
  10. Grim, J., Haindl, M., Somol, P., Pudil, P., A Subspace approach to texture modelling by using Gaussian mixtures, In: Proc. 18th Internat. Conference on Pattern Recognition. ICPR 2006 (Y. Y. Tang, S. P. Wang, D. S. Yeung, H. Yan, and G. Lorette, eds.), Vol. II, IEEE Computer Society, Los Alamitos 2006, pp. 235–238. (2006) 
  11. Haindl, M., Texture synthesis, Quarterly 4 (1991), 4, 305–331. (1991) Zbl0757.62051
  12. Haindl, M., Grim, J., Pudil, P., Kudo, M., A hybrid btf model based on gaussian mixtures, In: Texture 2005. The 4th Internat. Workshop on Texture Analysis and Synthesis in Conjunction with ICCV2005 (M. Chantler and O. Drbohlav, eds.), Beijing 2005, Heriot-Watt University & IEEE, pp. 95–100. (2005) 
  13. Haindl, M., Grim, J., Somol, P., Pudil, P., Kudo, M., A Gaussian mixture-based colour texture model, In: Proc. 17th IAPR Internat. Conference on Pattern Recognition (J. Kittler, M. Petrou, and M. Nixon, eds.), Vol. III, Los Alamitos 2004, IEEE, pp. 177–180. (2004) 
  14. Haindl, M., Havlíček, V., Multiresolution colour texture synthesis, In: Proc. 7th Internat. Workshop on Robotics in Alpe-Adria-Danube Region (K. Dobrovodský, ed.), Bratislava 1998. ASCO Art, pp. 297–302. (1998) 
  15. Haindl, M., Havlíček, V., A multiresolution causal colour texture model, In: Advances in Pattern Recognition (F. J. Ferri, J. M. Inesta, A. Amin, and P. Pudil, eds.), Lecture Notes in Computer Science 1876, Chapter 1, Springer-Verlag, Berlin 2000, pp. pages 114 –122. (1876) 
  16. Haindl, Michal, Filip, Jiří, 10.1109/TPAMI.2007.1139, IEEE Trans. Pattern Analysis and Machine Intelligence 29(2007), 10, 1859–1865. (2007) DOI10.1109/TPAMI.2007.1139
  17. Haindl, Michal, Hatka, Martin, Near-regular texture synthesis, In: Computer Analysis of Images and Patterns (X. Jiang and N. Petkov, eds.), Lecture Notes in Computer Sci. 5720, Springer 2009, pp. 1138–1145. (2009) 
  18. Haindl, Michal, Havlíček, Vojtěch, Grim, Jiří, Colour texture representation based on multivariate bernoulli mixtures, In: 10th Internat. Conference on Information Sciences, Signal Processing and their Applications (B. Boashash, R. Hamila, S. Hussain Shaikh Salleh, and S. Abd Rahman Abu Bakar, eds.), Kuala Lumpur 2010. IEEE, pp. 578–581. (2010) 
  19. Haindl, Michal, Havlíček, Vojtěch, Grim, Jiří, 10.1007/978-3-540-85920-8_82, Lecture Notes in Computer Sci. 5197 (2008), 675–682. (2008) DOI10.1007/978-3-540-85920-8_82
  20. Heeger, D. J., Bergen, J. R., Pyramid based texture analysis/synthesis, In: ACM SIGGRAPH 95, ACM Press 1995, pp. 229–238. (1995) 
  21. Koudelka, M. L., Magda, S., Belhumeur, P. N., Kriegman, D. J., Acquisition, compression, and synthesis of bidirectional texture functions, In: Texture 2003: Third Internat. Workshop on Texture Analysis and Synthesis, Nice 2003, pp. 59–64. (2003) 
  22. Nicodemus, F. E., Richmond, J. C., Hsia, J. J., Ginsburg, I. W., Limperis, T., Geometrical Considerations and Nomenclature for Feflectance, NBS Monograph No. 160, National Bureau of Standards, U.S. Department of Commerce, Washington, D. C. 1977. (1977) 
  23. Paget, R., Longstaff, I. D., 10.1109/83.679446, IEEE Trans. Image Processing 7 (1998), 8, 925–932. (1998) DOI10.1109/83.679446
  24. Sattler, M., Sarlette, R., Klein, R., Efficient and realistic visualization of cloth, In: Eurographics Symposium on Rendering 2003. (2003) 
  25. Xu, Y., Guo, B., Shum, H., Chaos Mosaic: Fast and Memory Efficient Texture Synthesis, Technical Report MSR-TR-2000-32, Redmont 2000. (2000) 
  26. Zelinka, S., Garland, M., Towards real-time texture synthesis with the jump map, In: 13th European Workshop on Rendering 2002, pp. 99–104. (2002) 
  27. Zhu, S. C., Liu, X. W., Wu, Y. N., 10.1109/34.862195, IEEE Trans. Pattern Analysis and Machine Intelligence 22 (2000), 6, 554–569. (2000) DOI10.1109/34.862195

NotesEmbed ?

top

You must be logged in to post comments.

To embed these notes on your page include the following JavaScript code on your page where you want the notes to appear.

Only the controls for the widget will be shown in your chosen language. Notes will be shown in their authored language.

Tells the widget how many notes to show per page. You can cycle through additional notes using the next and previous controls.

    
                

                
Note: Best practice suggests putting the JavaScript code just before the closing </body> tag.