This Article 
 Bibliographic References 
 Add to: 
Generalizing Swendsen-Wang to Sampling Arbitrary Posterior Probabilities
August 2005 (vol. 27 no. 8)
pp. 1239-1253
Many vision tasks can be formulated as graph partition problems that minimize energy functions. For such problems, the Gibbs sampler [9] provides a general solution but is very slow, while other methods, such as Ncut [24] and graph cuts [4], [22], are computationally effective but only work for specific energy forms [17] and are not generally applicable. In this paper, we present a new inference algorithm that generalizes the Swendsen-Wang method [25] to arbitrary probabilities defined on graph partitions. We begin by computing graph edge weights, based on local image features. Then, the algorithm iterates two steps. 1) Graph clustering: It forms connected components by cutting the edges probabilistically based on their weights. 2) Graph relabeling: It selects one connected component and flips probabilistically, the coloring of all vertices in the component simultaneously. Thus, it realizes the split, merge, and regrouping of a "chunk” of the graph, in contrast to Gibbs sampler that flips a single vertex. We prove that this algorithm simulates ergodic and reversible Markov chain jumps in the space of graph partitions and is applicable to arbitrary posterior probabilities or energy functions defined on graphs. We demonstrate the algorithm on two typical problems in computer vision—image segmentation and stereo vision. Experimentally, we show that it is 100-400 times faster in CPU time than the classical Gibbs sampler and 20-40 times faster then the DDMCMC segmentation algorithm [27]. For stereo, we compare performance with graph cuts and belief propagation. We also show that our algorithm can automatically infer generative models and obtain satisfactory results (better than the graphic cuts or belief propagation) in the same amount of time.

[1] A. Barbu and S.C. Zhu, “Graph Partition by Swendsen-Wang Cuts,” Proc. Int'l Conf. Computer Vision, 2003.
[2] A. Barbu and S.C. Zhu, “Multigrid and Multi-Level Swendsen-Wang Cuts for Hierarchic Graph Partition,” Proc. IEEE Conf. Computer Vision and Pattern Recognition, 2004.
[3] S.A. Barker, A.C. Kokaram, and P.J. Rayner, “Unsupervised Segmentation of Images,” Proc. SPIE Conf. Bayesian Inference for Inverse Problems, pp. 200-211, July 1998.
[4] Y. Boykov, O. Veksler, and R. Zabih, “Fast Approximate Energy Minimization via Graph Cuts,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 23, no. 11, pp. 1222-1239, Nov. 2001.
[5] C. Cooper and A. Frieze, “Mixing Properties of the Swendsen-Wang Process in Classes of Graphs,” Random Structures and Algorithms, vol. 15, no. 3-4, pp. 242-261, 1999.
[6] R.G. Edwards and A.D. Sokal, “Generalization of the Fortuin-Kasteleyn-Swendsen-Wang Representation and Monte Carlo Algorithm,” Physical Rev. Letters, vol. 38, pp. 2009-2012, 1988.
[7] C. Fox and G.K. Nicholls, “Exact MAP States and Expectations from Perfect Sampling,” Proc. 20th Int'l Workshop Bayesian Inference and Maximum Entropy Methods, 2000.
[8] Y. Gdalyahu, D. Weinshall, M. Werman, “Self-Organization in Vision: Stochastic Clustering for Image Segmentation, Perceptual Grouping and Image Database,” IEEE. Trans. Pattern Analysis and Machine Intelligence, vol. 23, no. 10, pp. 1053-1074, Oct. 2001.
[9] S. Geman and D. Geman, “Stochastic Relaxation, Gibbs Distributions, and the Bayesian Restoration of Images,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 6, pp. 721-741, 1984.
[10] V. Gore and A. Sinclair, “The Swendsen-Wang Process Does Not Always Mix Rapidly,” J. Statistical Physics, vol. 97, no. 1-2, pp. 67-86, 1999.
[11] W.K. Hastings, “Monte Carlo Sampling Methods Using Markov Chains and Their Applications,” Biometrika, vol. 57, pp. 97-109, 1970.
[12] D. Higdon, “Auxiliary Variable Methods for Markov Chain Monte Carlo Simulations,” preprint of the Inst. Statistics and Decision Science, 1996.
[13] T. Hofmann and J.M. Buhmann, “Pairwise Data Clustering by Deterministic Annealing,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 19, no. 1, pp. 1-14, Jan. 1997.
[14] M. Huber, “A Bounding Chain for Swendsen-Wang,” Random Structures and Algorithms, vol 22, no 1, pp. 43-59, 2002.
[15] A.K. Jain and R. Dubes, Algorithms for Clustering Data. Englewood Cliffs, N.J.: Prentice Hall, 1988.
[16] S. Kirkpatrick, C. Gelatt, and M. Vecchi, “Optimization by Simulated Annealing,” Science, vol. 220, no. 4598, pp. 671-680, 1983.
[17] V. Kolmogorov and R. Zabih, “What Energy Functions Can Be Minimized via Graph Cuts?” Proc. European Conf. Computer Vision, vol. 3, pp. 65-81, 2002.
[18] N. Metropolis, A. Rosenbluth, M. Rosenbluth, A. Teller, and E. Teller, “Equations of the State Calculations by Fast Computing Machines,” J. Chemical Physics, vol. 21, pp. 1087-1091, 1953.
[19] R.B. Potts, “Some Generalized Order-Disorder Transformations,” Proc. Cambridge Philosophic Soc., vol. 48, pp. 106-109, 1953.
[20] J.G. Propp and D.B. Wilson, “Exact Sampling with Coupled Markov Chains and Applications to Statistical Mechanics,” Random Structures and Algorithms, vol. 9, no. 1-2, pp. 223-252, 1996.
[21] J. Puzicha, T. Hofmann, and J.M. Buhmann, “A Theory of Proximity Based Clustering: Structure Detection by Optimization,” Pattern Recognition, vol. 33, no. 4, pp. 617-634, 1999.
[22] S. Roy and I. Cox, “A Maximum-Flow Formulation of the n-Camera Stereo Correspondence Problem,” Proc. Int'l Conf. Computer Vision, 1998.
[23] D. Scharstein and R. Szeliski, “A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms,” Int'l J. Computer Vision, 2002.
[24] J. Shi and J. Malik, “Normalized Cuts and Image Segmentation,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 22, no 8, pp. 888-905, Aug. 2000.
[25] R.H. Swendsen and J.S. Wang, “Nonuniversal Critical Dynamics in Monte Carlo Simulations,” Physical Rev. Letters, vol. 58, no. 2, pp. 86-88, 1987.
[26] M.F. Tappen and W.T. Freeman, “Comparison of Graph Cuts with Belief Propagation for Stereo, Using Identical MRF Parameters,” Proc. Int'l Conf. Computer Vision, 2003.
[27] Z.W. Tu and S.C. Zhu, “Image Segmentation by Data-Driven Markov Chain Monte Carlo,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 24, no. 5, pp. 657-673, May 2002.
[28] Z.W. Tu and S.C. Zhu, “Parsing Images into Region, Curves, and Curve Processes,” Proc. European Conf. Computer Vision, 2002.
[29] J. Wang et al. “Relationship between Ventral Stream for Object Recognition and Dorsal Stream for Spatial Vision: An FMRI and ERP Study,” Human Brain Mapping, vol. 8, pp. 170-181, 1999.
[30] G. Winkler, Image Analysis, Random Fields and Dynamic Monte Carlo Methods, pp. 171-173, Springer-Verlag, 1995.
[31] U. Wolff, “Collective Monte Carlo Updating for Spin Systems,” Physical Rev. Letters, vol. 62, no. 4, pp. 361-364, 1989.
[32] Z. Wu and R. Leahy, “An Optimal Graph Theoretic Approach to Data Clustering: Theory and Application to Image Segmentation,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 15, no. 11, pp. 1101-1113, Nov. 1993.
[33] J.S. Yedidia, W.T. Freeman, and Y. Weiss, “Generalized Belief Propagation,” MERL Report TR-2000-26, 2000.
[34] S.C. Zhu and A.L. Yuille, “Region Competition: Unifying Snake/Balloon, Region Growing and Bayes/MDL/Energy for Multi-Band Image Segmentation,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 18, no. 9, pp. 884-900, Sept. 1996.

Index Terms:
Index Terms- Swendsen-Wang, cluster sampling, Markov chain Monte Carlo, Bayesian inference, image segmentation, stereo matching.
Adrian Barbu, Song-Chun Zhu, "Generalizing Swendsen-Wang to Sampling Arbitrary Posterior Probabilities," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 27, no. 8, pp. 1239-1253, Aug. 2005, doi:10.1109/TPAMI.2005.161
Usage of this product signifies your acceptance of the Terms of Use.