Advertisement

On the Link between Gaussian Homotopy Continuation and Convex Envelopes

  • Hossein Mobahi
  • John W. FisherIII
Part of the Lecture Notes in Computer Science book series (LNCS, volume 8932)

Abstract

The continuation method is a popular heuristic in computer vision for nonconvex optimization. The idea is to start from a simplified problem and gradually deform it to the actual task while tracking the solution. It was first used in computer vision under the name of graduated nonconvexity. Since then, it has been utilized explicitly or implicitly in various applications. In fact, state-of-the-art optical flow and shape estimation rely on a form of continuation. Despite its empirical success, there is little theoretical understanding of this method. This work provides some novel insights into this technique. Specifically, there are many ways to choose the initial problem and many ways to progressively deform it to the original task. However, here we show that when this process is constructed by Gaussian smoothing, it is optimal in a specific sense. In fact, we prove that Gaussian smoothing emerges from the best affine approximation to Vese’s nonlinear PDE. The latter PDE evolves any function to its convex envelope, hence providing the optimal convexification.

Keywords

Continuation Method Diffusion Equation Nonconvex Optimization Vese’s PDE 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Balzer, J., Mörwald, T.: Isogeometric finite-elements methods and variational reconstruction tasks in vision - a perfect match. In: CVPR (2012)Google Scholar
  2. 2.
    Barron, J.: Shapes, Paint, and Light. Ph.D. thesis, EECS Department, University of California, Berkeley (August 2013)Google Scholar
  3. 3.
    Bengio, Y., Louradour, J., Collobert, R., Weston, J.: Curriculum learning. In: ICML (2009)Google Scholar
  4. 4.
    Bengio, Y.: Learning Deep Architectures for AI. Now Publishers Inc. (2009)Google Scholar
  5. 5.
    Black, M.J., Rangarajan, A.: On the unification of line processes, outlier rejection, and robust statistics with applications in early vision. International Journal of Computer Vision 19(1), 57–91 (1996)CrossRefGoogle Scholar
  6. 6.
    Blake, A., Zisserman, A.: Visual Reconstruction. MIT Press (1987)Google Scholar
  7. 7.
    Blake, A.: Comparison of the efficiency of deterministic and stochastic algorithms for visual reconstruction. IEEE PAMI 11(1), 2–12 (1989)CrossRefzbMATHMathSciNetGoogle Scholar
  8. 8.
    Boccuto, A., Discepoli, M., Gerace, I., Pucci, P.: A gnc algorithm for deblurring images with interacting discontinuities (2002)Google Scholar
  9. 9.
    Brox, T.: From pixels to regions: partial differential equations in image analysis. Ph.D. thesis, Saarland University, Germany (April 2005)Google Scholar
  10. 10.
    Brox, T., Malik, J.: Large displacement optical flow: Descriptor matching in variational motion estimation. IEEE Trans. Pattern Anal. Mach. Intell. 33(3), 500–513 (2011)CrossRefGoogle Scholar
  11. 11.
    Burgers, J.M.: The nonlinear diffusion equation: asymptotic solutions and statistical problems. D. Reidel Pub. Co. (1974)Google Scholar
  12. 12.
    Chapelle, O., Chi, M., Zien, A.: A continuation method for semi-supervised svms. pp. 185–192. ICML 2006 (2006)Google Scholar
  13. 13.
    Chapelle, O., Sindhwani, V., Keerthi, S.S.: Optimization techniques for semi-supervised support vector machines. J. Mach. Learn. Res. 9, 203–233 (2008)zbMATHGoogle Scholar
  14. 14.
    Chapelle, O., Wu, M.: Gradient descent optimization of smoothed information retrieval metrics. Inf. Retr. 13(3), 216–235 (2010)CrossRefGoogle Scholar
  15. 15.
    Chaudhuri, S., Clochard, M., Solar-Lezama, A.: Bridging boolean and quantitative synthesis using smoothed proof search. SIGPLAN Not 49(1), 207–220 (2014)Google Scholar
  16. 16.
    Chaudhuri, S., Solar-Lezama, A.: Smooth interpretation. In: PLDI. pp. 279–291. ACM (2010)Google Scholar
  17. 17.
    Chaudhuri, S., Solar-Lezama, A.: Smoothing a program soundly and robustly. In: Gopalakrishnan, G., Qadeer, S. (eds.) CAV 2011. LNCS, vol. 6806, pp. 277–292. Springer, Heidelberg (2011)CrossRefGoogle Scholar
  18. 18.
    Cohen, L.D., Gorre, A.: Snakes: Sur la convexite de la fonctionnelle denergie (1995)Google Scholar
  19. 19.
    Coupé, P., Manjón, J.V., Chamberland, M., Descoteaux, M., Hiba, B.: Collaborative patch-based super-resolution for diffusion-weighted images. NeuroImage 83, 245–261 (2013)CrossRefGoogle Scholar
  20. 20.
    Dai, Z., Lücke, J.: Unsupervised learning of translation invariant occlusive components. In: CVPR, pp. 2400–2407 (2012)Google Scholar
  21. 21.
    Dhillon, P.S., Keerthi, S.S., Bellare, K., Chapelle, O., Sundararajan, S.: Deterministic annealing for semi-supervised structured output learning. In: AISTATS 2012, vol. 15 (2012)Google Scholar
  22. 22.
    Dufour, R.M., Miller, E.L., Galatsanos, N.P.: Template matching based object recognition with unknown geometric parameters. IEEE Transactions on Image Processing 11(12), 1385–1396 (2002)CrossRefMathSciNetGoogle Scholar
  23. 23.
    Dvijotham, K., Fazel, M., Todorov, E.: Universal convexification via risk-aversion. In: UAI (2014)Google Scholar
  24. 24.
    Erhan, D., Manzagol, P.A., Bengio, Y., Bengio, S., Vincent, P.: The difficulty of training deep architectures and the effect of unsupervised pre-training. In: AISTATS, pp. 153–160 (2009)Google Scholar
  25. 25.
    Felzenszwalb, P.F., Girshick, R.B., McAllester, D.A., Ramanan, D.: Object detection with discriminatively trained part-based models. IEEE Trans. Pattern Anal. Mach. Intell. 32(9), 1627–1645 (2010)CrossRefGoogle Scholar
  26. 26.
    Frank, M., Streich, A.P., Basin, D., Buhmann, J.M.: Multi-assignment clustering for boolean data. J. Mach. Learn. Res. 13(1), 459–489 (2012)zbMATHMathSciNetGoogle Scholar
  27. 27.
    Fua, P., Leclerc, Y.: Object-centered surface reconstruction: combining multi-image stereo shading. International Journal on Computer Vision 16(1), 35–56 (1995)CrossRefGoogle Scholar
  28. 28.
    Gehler, P., Chapelle, O.: Deterministic annealing for multiple-instance learning. In: AISTATS 2007, pp. 123–130. Microtome, Brookline (2007)Google Scholar
  29. 29.
    Geiger, D., Girosi, F.: Coupled markov random fields and mean field theory. In: NIPS, pp. 660–667. Morgan Kaufmann (1989)Google Scholar
  30. 30.
    Geiger, D., Yuille, A.L.: A common framework for image segmentation. International Journal of Computer Vision 6(3), 227–243 (1991)CrossRefGoogle Scholar
  31. 31.
    Gold, S., Rangarajan, A.: A graduated assignment algorithm for graph matching. IEEE PAMI 18, 377–388 (1996)CrossRefGoogle Scholar
  32. 32.
    Gold, S., Rangarajan, A., Mjolsness, E.: Learning with preknowledge: Clustering with point and graph matching distance measures. In: NIPS, pp. 713–720 (1994)Google Scholar
  33. 33.
    Held, D., Levinson, J., Thrun, S., Savarese, S.: Combining 3d shape, color, and motion for robust anytime tracking. In: RSS, Berkeley, USA (July 2014)Google Scholar
  34. 34.
    Hinton, G.E., Osindero, S., Teh, Y.W.: A fast learning algorithm for deep belief networks. Neural Computation 18(7), 1527–1554 (2006)CrossRefzbMATHMathSciNetGoogle Scholar
  35. 35.
    Hong, B.W., Lu, Z., Sundaramoorthi, G.: A new model and simple algorithms for multi-label mumford-shah problems. In: CVPR (June 2013)Google Scholar
  36. 36.
    Kim, J., Liu, C., Sha, F., Grauman, K.: Deformable spatial pyramid matching for fast dense correspondences. In: CVPR, pp. 2307–2314. IEEE (2013)Google Scholar
  37. 37.
    Kim, M., Torre, F.D.: Gaussian processes multiple instance learning, pp. 535–542 (2010)Google Scholar
  38. 38.
    Kosowsky, J.J., Yuille, A.L.: The invisible hand algorithm: Solving the assignment problem with statistical physics. Neural Networks 7(3), 477–490 (1994)CrossRefzbMATHGoogle Scholar
  39. 39.
    Leich, A., Junghans, M., Jentschel, H.J.: Hough transform with GNC. 12th European Signal Processing Conference (EUSIPCO, 2004)Google Scholar
  40. 40.
    Leordeanu, M., Hebert, M.: Smoothing-based optimization. In: CVPR (2008)Google Scholar
  41. 41.
    Li, X.: Fine-granularity and spatially-adaptive regularization for projection-based image deblurring. IEEE Transactions on Image Processing 20(4), 971–983 (2011)CrossRefMathSciNetGoogle Scholar
  42. 42.
    Liu, Z., Qiao, H., Xu, L.: An extended path following algorithm for graph-matching problem. IEEE Trans. Pattern Anal. Mach. Intell. 34(7), 1451–1456 (2012)CrossRefGoogle Scholar
  43. 43.
    Loog, M., Duistermaat, J.J., Florack, L.M.J.: On the behavior of spatial critical points under gaussian blurring (A folklore theorem and scale-space constraints). In: Kerckhove, M. (ed.) Scale-Space 2001. LNCS, vol. 2106, pp. 183–192. Springer, Heidelberg (2001)CrossRefGoogle Scholar
  44. 44.
    Mairal, J., Bach, F., Ponce, J., Sapiro, G.: ICML, vol. 382, p. 87. ACM (2009)Google Scholar
  45. 45.
    Malek-Mohammadi, M., Babaie-Zadeh, M., Amini, A., Jutten, C.: Recovery of low-rank matrices under affine constraints via a smoothed rank function. IEEE Transactions on Signal Processing 62(4), 981–992 (2014)CrossRefMathSciNetGoogle Scholar
  46. 46.
    Mobahi, H., Rao, S., Ma, Y.: Data-driven image completion by image patch subspaces. In: Picture Coding Symposium (2009)Google Scholar
  47. 47.
    Mobahi, H., Ma, Y., Zitnick, L.: Seeing through the Blur. In: CVPR (2012)Google Scholar
  48. 48.
    Mohimani, G.H., Babaie-Zadeh, M., Gorodnitsky, I., Jutten, C.: Sparse recovery using smoothed ℓ0 (sl0): Convergence analysis. CoRR abs/1001.5073 (2010)Google Scholar
  49. 49.
    Mumford, D., Shah, J.: Optimal approximations by piecewise smooth functions and associated variational problems. Comm. Pure Appl. Math. 42(5), 577–685 (1989)CrossRefzbMATHMathSciNetGoogle Scholar
  50. 50.
    Nielsen, M.: Graduated non-convexity by smoothness focusing. In: Proceedings of the British Machine Vision Conference, pp. 60.1–60.10. BMVA Press (1993)Google Scholar
  51. 51.
    Nikolova, M., Ng, M.K., Tam, C.P.: Fast nonconvex nonsmooth minimization methods for image restoration and reconstruction. Trans. Img. Proc. 19(12), 3073–3088 (2010)CrossRefMathSciNetGoogle Scholar
  52. 52.
    Pretto, A., Soatto, S., Menegatti, E.: Scalable dense large-scale mapping and navigation. In: Proc. of: Workshop on Omnidirectional Robot Vision, ICRA (2010)Google Scholar
  53. 53.
    Price, B.L., Morse, B.S., Cohen, S.: Simultaneous foreground, background, and alpha estimation for image matting. In: CVPR, pp. 2157–2164. IEEE (2010)Google Scholar
  54. 54.
    Rangarajan, A., Chellappa, R.: Generalized graduated nonconvexity algorithm for maximum a posteriori image estimation, pp. II:127–II:133 (1990)Google Scholar
  55. 55.
    Rose, K., Gurewitz, E., Fox, G.: A deterministic annealing approach to clustering. Pattern Recognition Letters 11(9), 589–594 (1990)CrossRefzbMATHGoogle Scholar
  56. 56.
    Rossi, F., Villa-Vialaneix, N.: Optimizing an organized modularity measure for topographic graph clustering: A deterministic annealing approach. Neurocomputing 73(7-9), 1142–1163 (2010)CrossRefGoogle Scholar
  57. 57.
    Saragih, J.: Deformable face alignment via local measurements and global constraints, pp. 187–207. Springer, Heidelberg (2013)Google Scholar
  58. 58.
    Shroff, N., Turaga, P.K., Chellappa, R.: Manifold precis: An annealing technique for diverse sampling of manifolds. In: NIPS, pp. 154–162 (2011)Google Scholar
  59. 59.
    Sindhwani, V., Keerthi, S.S., Chapelle, O.: Deterministic annealing for semi-supervised kernel machines. In: ICML 2006, pp. 841–848. ACM, New York (2006)Google Scholar
  60. 60.
    Smith, N.A., Eisner, J.: Annealing techniques for unsupervised statistical language learning. In: ACL, Barcelona, Spain, pp. 486–493 (July 2004)Google Scholar
  61. 61.
    Stoll, M., Volz, S., Bruhn, A.: Joint trilateral filtering for multiframe optical flow. In: ICIP, pp. 3845–3849 (2013)Google Scholar
  62. 62.
    Sun, D., Roth, S., Black, M.J.: Secrets of optical flow estimation and their principles. In: CVPR, pp. 2432–2439. IEEE (2010)Google Scholar
  63. 63.
    Terzopoulos, D.: The computation of visible-surface representations. IEEE Trans. Pattern Anal. Mach. Intell. 10(4), 417–438 (1988)CrossRefzbMATHGoogle Scholar
  64. 64.
    Tirthapura, S., Sharvit, D., Klein, P., Kimia, B.: Indexing based on edit-distance matching of shape graphs. In: SPIE International Symposium on Voice, Video, and Data Communications, pp. 25–36 (1998)Google Scholar
  65. 65.
    Trzasko, J., Manduca, A.: Highly undersampled magnetic resonance image reconstruction via homotopic ℓ0 -minimization. IEEE Trans. Med. Imaging 28(1), 106–121 (2009)CrossRefGoogle Scholar
  66. 66.
    Vese, L.: A method to convexify functions via curve evolution. Commun. Partial Differ. Equations 24(9-10), 1573–1591 (1999)CrossRefzbMATHMathSciNetGoogle Scholar
  67. 67.
    Vural, E., Frossard, P.: Analysis of descent-based image registration. SIAM J. Imaging Sciences 6(4), 2310–2349 (2013)CrossRefzbMATHMathSciNetGoogle Scholar
  68. 68.
    Widder, D.V.: The Heat Equation. Academic Press (1975)Google Scholar
  69. 69.
    Wright, J., Yang, A.Y., Ganesh, A., Sastry, S.S., Ma, Y.: Robust face recognition via sparse representation. IEEE Trans. Pattern Anal. Mach. Intell. 31(2), 210–227 (2009)CrossRefGoogle Scholar
  70. 70.
    Wu, Z., Tan, P.: Calibrating photometric stereo by holistic reflectance symmetry analysis, pp. 1498–1505. IEEE (2013)Google Scholar
  71. 71.
    Wu, Z.: The Effective Energy Transformation Scheme as a Special Continuation Approach to Global Optimization with Application to Molecular Conformation. SIAM J. on Optimization 6, 748–768 (1996)CrossRefzbMATHGoogle Scholar
  72. 72.
    Yuille, A.: Energy Functions for Early Vision and Analog Networks. A.I. memo, Defense Technical Information Center (1987)Google Scholar
  73. 73.
    Yuille, A.L.: Generalized deformable models, statistical physics, and matching problems. Neural Computation 2, 1–24 (1990)CrossRefGoogle Scholar
  74. 74.
    Yuille, A., Geiger, D., Bulthoff, H.: Stereo integration, mean field theory and psychophysics. In: Faugeras, O. (ed.) ECCV 1990. LNCS, vol. 427, pp. 71–82. Springer, Heidelberg (1990)CrossRefGoogle Scholar
  75. 75.
    Yuille, A.L., Peterson, C., Honda, K.: Deformable templates, robust statistics, and hough transforms, San Diego, CA, pp. 166–174. International Society for Optics and Photonics (1991)Google Scholar
  76. 76.
    Yuille, A.L., Stolorz, P.E., Utans, J.: Statistical physics, mixtures of distributions, and the em algorithm. Neural Computation 6(2), 334–340 (1994)CrossRefGoogle Scholar
  77. 77.
    Zaslavskiy, M., Bach, F., Vert, J.P.: A path following algorithm for the graph matching problem. IEEE PAMI (2009)Google Scholar
  78. 78.
    Zerubia, J., Chellappa, R.: Mean field annealing using compound gauss-markov random fields for edge detection and image estimation. IEEE Transactions on Neural Networks 4(4), 703–709 (1993)CrossRefGoogle Scholar

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  • Hossein Mobahi
    • 1
  • John W. FisherIII
    • 1
  1. 1.Computer Science and Artificial Intelligence Lab. (CSAIL)Massachusetts Institute of Technology (MIT)USA

Personalised recommendations