This Article 
 Bibliographic References 
 Add to: 
Prior Learning and Gibbs Reaction-Diffusion
November 1997 (vol. 19 no. 11)
pp. 1236-1250

Abstract—This article addresses two important themes in early visual computation: First, it presents a novel theory for learning the universal statistics of natural images—a prior model for typical cluttered scenes of the world—from a set of natural images, and, second, it proposes a general framework of designing reaction-diffusion equations for image processing. We start by studying the statistics of natural images including the scale invariant properties, then generic prior models were learned to duplicate the observed statistics, based on the minimax entropy theory studied in two previous papers. The resulting Gibbs distributions have potentials of the form $U\left( {{\schmi{\bf I}};\,\Lambda ,\,S} \right)=\sum\nolimits_{\alpha =1}^K {\sum\nolimits_{ \left( {x,y} \right)} {\lambda ^{\left( \alpha \right)}}}\left( {\left( {F^{\left( \alpha \right)}*{\schmi{\bf I}}} \right)\left( {x,y} \right)} \right)$ with S = {F(1), F(2), ..., F(K)} being a set of filters and Λ = {λ(1)(), λ(2)(), ..., λ(K)()} the potential functions. The learned Gibbs distributions confirm and improve the form of existing prior models such as line-process, but, in contrast to all previous models, inverted potentials (i.e., λ(x) decreasing as a function of |x|) were found to be necessary. We find that the partial differential equations given by gradient descent on U(I; Λ, S) are essentially reaction-diffusion equations, where the usual energy terms produce anisotropic diffusion, while the inverted energy terms produce reaction associated with pattern formation, enhancing preferred image features. We illustrate how these models can be used for texture pattern rendering, denoising, image enhancement, and clutter removal by careful choice of both prior and data models of this type, incorporating the appropriate features.

[1] A. Berger, V. Della Pietra, and S. Della Pietra, "A Maximum Entropy Approach to Natural Language Processing," Computational Linguistics, vol. 22, no. 1, 1996.
[2] M. Black, G. Sapiro, D. Marimont, and D. Heeger, “Robust Anisotropic Diffusion,” IEEE Trans. Image Processing, vol. 7, no. 3, pp. 421-432, 1998.
[3] M.J. Black and A. Rangarajan, "On the Unification of Line Processes, Outlier Rejection, and Robust Statistics With Applications in Early Vision," Int'l J. Computer Vision, vol. 19, no. 1, 1996.
[4] A. Blake and A. Zisserman, Visual Reconstruction.Cambridge, Mass.: MIT Press, 1987.
[5] J. Daugman, "Uncertainty Relation for Resolution in Space, Spatial Frequency, and Orientation Optimized by Two-Dimensional Visual Cortical Filters," J. Optical Soc. America, vol. 2, no. 7, pp. 1,160-1,169, 1985.
[6] D.J. Field, "Relations Between the Statistics of Natural Images and the Response Properties of Cortical Cells," J. Optical Soc. America, A, vol. 4, no. 12, 1987.
[7] D. Gabor, "Theory of Communication," IEE Proc., vol. 93, no. 26, 1946.
[8] S.B. Gelfand and S.K. Mitter, "On Sampling Methods and Annealing Algorithms," Markov Random Fields—Theory and Applications.New York: Academic Press, 1993.
[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, no. 7, pp. 721-741, July 1984.
[10] S. Geman and C. Hwang, "Diffusion for Global Optimization," SIAM J. Control and Optimization, vol. 24, no. 5, 1986.
[11] D. Geman and G. Reynolds, "Constrained Restoration and the Recovery of Discontinuities," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, pp. 367-383, 1992.
[12] D. Geiger and F. Girosi,“Parallel and deterministic algorithms from MRFs: Surface reconstruction,” IEEE Transactions on PAMI, vol. 13, no. 5, pp. 401-412, May 1991.
[13] D. Geiger and A.L. Yuille, "A Common Framework for Image Segmentation," Int'l J. Computer Vision, vol. 6, no. 3, pp. 227-243, 1991.
[14] B. Gidas, "A Renormalization Group Approach to Image Processing Problems," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 11, pp. 164-180, 1989.
[15] P. Grindrod, The Theory and Applications of Reaction-Diffusion Equations.New York: Oxford Univ. Press, 1996.
[16] B. Kimia, A. Tannebaum, and S. Zucker, "Shapes, Shocks, and Deformations I: The Components of Two-Dimensional Shape and the Reaction-Diffusion Space," Int'l J. Computer Vision, vol. 15, pp. 189-224, 1995.
[17] S. Kullback and R.A. Leibler, "On Information and Sufficiency," Annual Math. Stat., vol. 22, pp. 79-86, 1951.
[18] J. Marroguin, S. Mitter, and T. Poggio, "Probabilistic Solution of Ill-Posed Problems in Computational Vision," J. Am. Statistical Assoc., vol. 82, no. 397, 1987.
[19] P. Meer, D. Mintz, D.Y. Kim, and A. Rosenfeld, "Robust Regression Methods for Computer Vision: A Review," Int'l J. Computer Vision, vol. 6, no. 1, 1991.
[20] D. Mumford and J. Shah, "Optimal Approximations by Piecewise Smooth Functions and Associated Variational Problems," Comm. Pure Applied Math., vol. 42, pp. 577-684, 1989.
[21] J.D. Murray, "A Pre-Pattern Formation Mechanism for Mammalian Coat Markings," J. Theoretical Biology, vol. 88, pp. 161-199, 1981.
[22] W. Niessen, B. Romeny, L. Florack, and M. Viergever, "A General Framework for Geometry-Driven Evolution Equations," Int'l J. Computer Vision, vol. 21, no. 3, pp. 187-205, 1997.
[23] M. Nitzberg and T. Shiota, “Nonlinear Image Filtering with Edge and Corner Enhancement,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 14, pp. 826-833, 1992.
[24] B.A. Olshausen and D.J. Field, "Natural Image Statistics and Efficient Coding," Proc. Workshop on Information Theory and the Brain, Sept. 1995.
[25] P. Perona and J. Malik, "Scale-Space and Edge Detection Using Anisotropic Diffusion," IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12, no. 7, pp. 629639, July 1990.
[26] T. Poggio, V. Torre, and C. Koch, "Computational Vision and Regularization Theory," Nature, vol. 317, pp. 314-319, 1985.
[27] T. Poggio and F. Girosi, Networks for Approximation and Learning Proc. IEEE, vol. 78, pp. 1481-1497, 1990.
[28] D.L. Ruderman and W. Bialek, "Statistics of Natural Images: Scaling in the Woods," Phys. Rev. Letter, vol. 73, pp. 814-817, 1994.
[29] A. Sherstinsky and R. Picard, "M-Lattice: From Morphogenesis to Image Processing," IEEE Trans. Image Processing, vol. 5, no. 7, July 1996.
[30] N.V. Swindale, "A Model for the Formation of Ocular Dominance Stripes," Proc. Royal Soc. London B, vol. 208, pp. 243-264, 1980.
[31] D. Terzopoulos, "Multilevel Computational Processes for Visual Surface Reconstruction," Computer Vision, Graphics, and Image Processing, vol. 24, pp. 52-96, 1983.
[32] A.N. Tikhonov and V.Y. Arsenin, Solutions of Ill-Posed Problems. V.H. Winston&Sons, 1977.
[33] A.M. Turing, "The Chemical Basis of Morphogenesis," Philosophy Trans. Royal Soc. London, vol. 237, no. B, pp. 37-72, 1952.
[34] G. Turk, "Generating Textures on Arbitrary Surfaces Using Reaction-Diffusion," Computer Graphics, vol. 25, no. 4, 1991.
[35] A.B. Watson, "Efficiency of Model Human Image Code," J. Optical Soc. America A, vol. 4, no. 12, 1987.
[36] K. Wilson, "The Renormalization Group: Critical Phenomena and the Knodo Problem," Rev. Mod. Phys., vol. 47, pp. 773-840, 1975.
[37] G. Winkler, Image Analysis, Random Fields and Dynamic Monte Carlo Methods.New York: Springer-Verlag, 1995.
[38] A. Witkin and M. Kass, "Reaction-Diffusion Textures," Computer Graphics, vol. 25, no. 4, 1991.
[39] S.C. Zhu and A. Yuille, “Region Competition: Unifying Snakes, Region Growing and Bayes/MDL for Multiband Image Segmentation,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 18, pp. 884-900, 1996.
[40] S.C. Zhu, Y. Wu, and D. Mumford, “Filters, Random Fields and Maximum Entropy (FRAME)—Towards a Unified Theory for Texture Modeling,” Int'l J. Computer Vision, vol. 27, no. 2, 1998.
[41] S.C. Zhu, Y.N. Wu, and D.B. Mumford, "Minimax Entropy Principle and Its Application to Texture Modeling," Neural Computation, vol. 9, no. 8, Nov. 1997.
[42] S.C. Zhu and D.B. Mumford, "Learning Generic Prior Models for Visual Computation," Harvard Robotics Lab, Technical Report TR-96-05 (a short version appeared in CVPR97).

Index Terms:
Visual learning, Gibbs distribution, reaction-diffusion, anisotropic diffusion, texture synthesis, clutter modeling, image restoration.
Song Chun Zhu, David Mumford, "Prior Learning and Gibbs Reaction-Diffusion," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 11, pp. 1236-1250, Nov. 1997, doi:10.1109/34.632983
Usage of this product signifies your acceptance of the Terms of Use.