A Comparative Study of Energy Minimization Methods for Markov Random Fields

  • Richard Szeliski
  • Ramin Zabih
  • Daniel Scharstein
  • Olga Veksler
  • Vladimir Kolmogorov
  • Aseem Agarwala
  • Marshall Tappen
  • Carsten Rother
Part of the Lecture Notes in Computer Science book series (LNCS, volume 3952)


One of the most exciting advances in early vision has been the development of efficient energy minimization algorithms. Many early vision tasks require labeling each pixel with some quantity such as depth or texture. While many such problems can be elegantly expressed in the language of Markov Random Fields (MRF’s), the resulting energy minimization problems were widely viewed as intractable. Recently, algorithms such as graph cuts and loopy belief propagation (LBP) have proven to be very powerful: for example, such methods form the basis for almost all the top-performing stereo methods. Unfortunately, most papers define their own energy function, which is minimized with a specific algorithm of their choice. As a result, the tradeoffs among different energy minimization algorithms are not well understood. In this paper we describe a set of energy minimization benchmarks, which we use to compare the solution quality and running time of several common energy minimization algorithms. We investigate three promising recent methods—graph cuts, LBP, and tree-reweighted message passing—as well as the well-known older iterated conditional modes (ICM) algorithm. Our benchmark problems are drawn from published energy functions used for stereo, image stitching and interactive segmentation. We also provide a general-purpose software interface that allows vision researchers to easily switch between optimization methods with minimal overhead. We expect that the availability of our benchmarks and interface will make it significantly easier for vision researchers to adopt the best method for their specific problems. Benchmarks, code, results and images are available at http://vision.middlebury.edu/MRF.


Energy Function Stereo Match IEEE Trans Pattern Anal Markov Random Energy Minimization Method 
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.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Besag, J.: On the statistical analysis of dirty pictures (with discussion). Journal of the Royal Statistical Society, Series B 48, 259–302 (1986)MathSciNetMATHGoogle Scholar
  2. 2.
    Geman, S., Geman, D.: Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. IEEE Trans Pattern Anal Mach Intell 6, 721–741 (1984)CrossRefMATHGoogle Scholar
  3. 3.
    Barnard, S.: Stochastic stereo matching over scale. Intern Journ Comp Vis 3, 17–32 (1989)CrossRefGoogle Scholar
  4. 4.
    Boykov, Y., Veksler, O., Zabih, R.: Fast approximate energy minimization via graph cuts. IEEE Trans Pattern Anal Mach Intell 23, 1222–1239 (2001)CrossRefGoogle Scholar
  5. 5.
    Kolmogorov, V., Zabih, R.: What energy functions can be minimized via graph cuts? IEEE Trans Pattern Anal Mach Intell 26, 147–159 (2004)CrossRefMATHGoogle Scholar
  6. 6.
    Yedidia, J.S., Freeman, W.T., Weiss, Y.: Generalized belief propagation. In: NIPS, pp. 689–695 (2000)Google Scholar
  7. 7.
    Scharstein, D., Szeliski, R.: A taxonomy and evaluation of dense two-frame stereo correspondence algorithms. Intern Journ Comp Vis 47, 7–42 (2002)CrossRefMATHGoogle Scholar
  8. 8.
    Geman, S., Graffigne, C.: Markov Random Field image models and their applications to computer vision. In: Intern Cong of Mathematicians, pp. 1496–1517 (1986)Google Scholar
  9. 9.
    Chou, P.B., Brown, C.M.: The theory and practice of Bayesian image labeling. Intern Journ Comp Vis 4, 185–210 (1990)CrossRefGoogle Scholar
  10. 10.
    Boykov, Y., Jolly, M.P.: Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images. In: ICCV, pp. I:105–112 (2001)Google Scholar
  11. 11.
    Rother, C., Kolmogorov, V., Blake, A.: “GrabCut” - interactive foreground extraction using iterated graph cuts. SIGGRAPH 23, 309–314 (2004)CrossRefGoogle Scholar
  12. 12.
    Agarwala, A., et al.: Interactive digital photomontage. SIGGRAPH 23, 292–300 (2004)Google Scholar
  13. 13.
    Tappen, M.F., Freeman, W.T.: Comparison of graph cuts with belief propagation for stereo, using identical MRF parameters. In: ICCV, pp. 900–907 (2003)Google Scholar
  14. 14.
    Boykov, Y., Kolmogorov, V.: An experimental comparison of min-cut/max-flow algorithms for energy minimization in vision. IEEE Trans Pattern Anal Mach Intell 26, 1124–1137 (2004)CrossRefMATHGoogle Scholar
  15. 15.
    Kolmogorov, V., Rother, C.: Comparison of energy minimization algorithms for highly connected graphs. In: Leonardis, A., Bischof, H., Pinz, A. (eds.) ECCV 2006. LNCS, vol. 3952, pp. 1–15. Springer, Heidelberg (2006)CrossRefGoogle Scholar
  16. 16.
    Li, S.: Markov Random Field Modeling in Computer Vision. Springer, Heidelberg (1995)CrossRefGoogle Scholar
  17. 17.
    Efros, A.A., Freeman, W.T.: Image quilting for texture synthesis and transfer. SIGGRAPH, 341–346 (2001)Google Scholar
  18. 18.
    Kwatra, V., Schodl, A., Essa, I., Turk, G., Bobick, A.: Graphcut textures: Image and video synthesis using graph cuts. SIGGRAPH (2003)Google Scholar
  19. 19.
    Greig, D., Porteous, B., Seheult, A.: Exact maximum a posteriori estimation for binary images. Journal of the Royal Statistical Society, Series B 51, 271–279 (1989)Google Scholar
  20. 20.
    Ishikawa, H.: Exact optimization for Markov Random Fields with convex priors. IEEE Trans Pattern Anal Mach Intell 25, 1333–1336 (2003)CrossRefGoogle Scholar
  21. 21.
    Hochbaum, D.S.: An efficient algorithm for image segmentation, Markov Random Fields and related problems. Journal of the ACM (JACM) 48, 686–701 (2001)MathSciNetCrossRefMATHGoogle Scholar
  22. 22.
    Felzenszwalb, P.F., Huttenlocher, D.P.: Efficient belief propagation for early vision. In: CVPR, pp. 261–268 (2004)Google Scholar
  23. 23.
    Kolmogorov, V.: Convergent tree-reweighted message passing for energy minimization. In: AISTATS (2005)Google Scholar
  24. 24.
    Rother, C., Kumar, S., Kolmogorov, V., Blake, A.: Digital tapestry. In: CVPR (2005)Google Scholar
  25. 25.
    Pearl, J.: Probabilistic reasoning in intelligent systems: networks of plausible inference. Morgan Kaufmann, San Francisco (1988)MATHGoogle Scholar
  26. 26.
    Agarwala, A., et al.: Panoramic video textures. SIGGRAPH 24, 821–827 (2005)CrossRefGoogle Scholar
  27. 27.
    Snow, D., Viola, P., Zabih, R.: Exact voxel occupancy with graph cuts. In: CVPR, pp. 345–352 (2000)Google Scholar
  28. 28.
    Buehler, C., Gortler, S.J., Cohen, M.F., McMillan, L.: Minimal surfaces for stereo. In: Heyden, A., Sparr, G., Nielsen, M., Johansen, P. (eds.) ECCV 2002. LNCS, vol. 2352, pp. 885–899. Springer, Heidelberg (2002)CrossRefGoogle Scholar
  29. 29.
    Kolmogorov, V., Criminisi, A., Blake, A., Cross, G., Rother, C.: Bi-layer segmentation of binocular stereo video. In: CVPR, pp. 407–414 (2005)Google Scholar
  30. 30.
    Wainwright, M.J., Jaakkola, T.S., Willsky, A.S.: MAP estimation via agreement on (hyper)trees: Message-passing and linear-programming approaches. IEEE Trans Info Theory 51 (2005)Google Scholar
  31. 31.
    Meltzer, T., Yanover, C., Weiss, Y.: Globally optimal solutions for energy minimization in stereo vision using reweighted belief propagation. In: ICCV (2005)Google Scholar
  32. 32.
    Komodakis, N., Tziritas, G.: A new framework for approximate labeling via graph cuts. In: ICCV (2005)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2006

Authors and Affiliations

  • Richard Szeliski
    • 1
  • Ramin Zabih
    • 2
  • Daniel Scharstein
    • 3
  • Olga Veksler
    • 4
  • Vladimir Kolmogorov
    • 5
  • Aseem Agarwala
    • 6
  • Marshall Tappen
    • 7
  • Carsten Rother
    • 1
  1. 1.Microsoft ResearchUSA
  2. 2.Cornell UniversityUSA
  3. 3.Middlebury CollegeUSA
  4. 4.University of Western OntarioCanada
  5. 5.University College LondonUK
  6. 6.University of WashingtonUSA
  7. 7.MITUSA

Personalised recommendations