[1]SAIKA Yohei,AOKI Toshizumi.Thermodynamicsinspired inverse halftoning via multiple halftone images[J].智能系统学报,2012,7(1):86-94.
SAIKA Yohei,AOKI Toshizumi.Thermodynamicsinspired inverse halftoning via multiple halftone images[J].CAAI Transactions on Intelligent Systems,2012,7(1):86-94.
点击复制
SAIKA Yohei,AOKI Toshizumi.Thermodynamicsinspired inverse halftoning via multiple halftone images[J].CAAI Transactions on Intelligent Systems,2012,7(1):86-94.
Thermodynamicsinspired inverse halftoning via multiple halftone images
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
7
期数:
2012年第1期
页码:
86-94
栏目:
学术论文—人工智能基础
出版日期:
2012-02-25
- Title:
- Thermodynamicsinspired inverse halftoning via multiple halftone images
- 文章编号:
- 1643-4785(2012)01-0086-09
- 分类号:
- TP39
- 文献标志码:
- A
- 摘要:
- Based on an analogy between thermodynamics and Bayesian inference, inverse halftoning was formulated using multiple halftone images based on Bayesian inference using the maximizer of the posterior marginal (MPM) estimate. Applying Monte Carlo simulation to a set of snapshots of the QIsing model, it was demonstrated that optimal performance is achieved around the Bayesoptimal condition within statistical uncertainty and that the performance of the Bayesoptimal solution is superior to that of the maximumaposteriori(MAP) estimation which is a deterministic limit of the MPM estimate. These properties were qualitatively confirmed by the meanfield theory using an infiniterange model established in statistical mechanics. Additionally, a practical and useful method was constructed using the statistical mechanical iterative method via the Bethe approximation. Numerical simulations for a 256grayscale standard image show that Bethe approximation works as good as the MPM estimation if the parameters are set appropriately.
- Abstract:
- Based on an analogy between thermodynamics and Bayesian inference, inverse halftoning was formulated using multiple halftone images based on Bayesian inference using the maximizer of the posterior marginal (MPM) estimate. Applying Monte Carlo simulation to a set of snapshots of the QIsing model, it was demonstrated that optimal performance is achieved around the Bayesoptimal condition within statistical uncertainty and that the performance of the Bayesoptimal solution is superior to that of the maximumaposteriori(MAP) estimation which is a deterministic limit of the MPM estimate. These properties were qualitatively confirmed by the meanfield theory using an infiniterange model established in statistical mechanics. Additionally, a practical and useful method was constructed using the statistical mechanical iterative method via the Bethe approximation. Numerical simulations for a 256grayscale standard image show that Bethe approximation works as good as the MPM estimation if the parameters are set appropriately.
- 参考文献/References:
-
［1］BESAG J. On the statistical analysis of dirty pictures［J］. Journal of the Royal Statistical Society—Series B: Statistical Methodological, 1986, 48(3): 259302.
［2］GONZALES R C, WOODS R C. Digital image processing［M］. Reading, USA: Addison Wiley, 1986.
［3］WINKLER G. Image analysis, random fields and dynamic Monte Carlo methods: a mathematical introduction［M］. Berlin, Germany: SpringerVerlag, 1995.
［4］NISHIMORI H. Statistical physics of spin glasses and information processing: an introduction［M］. Oxford, UK: Oxford Press, 2001.
［5］TANAKA K. Statisticalmechanical approach to image processing［J］. Journal of Physics A: Mathematical and General, 2002, 35(37): R31R150.
［6］ULICHNEY R. Digital halftoning［M］. Cambridge, USA: The MIT Press, 1987.
［7］BAYER B E. An optimum method for twolevel rendition of continuous tone pictures［C］//IEEE International Conference on Communications. New York, USA: IEEE Press, 1973: 1115.
［8］MICELI C M, PARKER K J. Inverse halftoning［J］. Journal of Electronic Imaging, 1992, 1(2): 143151.
［9］WONG P W. Inverse halftoning and kernel estimation for error diffusion［J］. IEEE Transactions on Image Processing, 1995, 4(4): 486498.
［10］STEVENSON R L. Inverse halftoning via MAP estimation［J］. IEEE Transactions on Image Processing, 1997, 6(4): 574583.
［11］SAIKA Y, INOUE J, TANAKA H, et al. Bayesoptimal solution to inverse halftoning based on statistical mechanics of the QIsing model［J］. Central European Journal of Physics, 2009, 7(3): 444456.
［12］TIPPING M E, BISHOP C M. Bayesian image superresolution［M］//BECKER S, THRUN S, OBERMAYER K. Advances in Neural Information Processing Systems. Cambridge, USA: The MIT Press, 2003: 12791286.
［13］KANEMURA A, MAEDA S, ISHII S. Superresolution with compound Markov random fields via the variational EM algorithm［J］. Neural Networks, 2009, 22(7): 10251034.
［14］DEMPSTER A P, LAIRD N M, RUBIN D B. Maximumlikelihood from incomplete data via the EM algorithm［J］. Journal of the Royal Statistical Society—Series B: Statistical Methodological, 1977, 39(1): 138.
［15］TANAKA K, INOUE J, TITTERINGTON D M. Probabilistic image processing by means of the Bethe approximation for the QIsing model［J］. Journal of Physics A: Mathematical and General, 2003, 36(43): 1102311035.
备注/Memo
Received Date: 2011-11-01.
?Corresponding Author: SAIKA Yohei.????????E-mail:yoheisaika@gmail.com.
About the authors：
SAIKA Yohei，received a B.S. degree in physics from Tokyo Science University, Tokyo, Japan in 1989, an M.S. degree in physics from Tokyo Institute of Technology in 1991, and a Ph.D. degree in physics from Tokyo Institute of Technology in 1995.
AOKI Toshizumi，received a B.S. degree in physics from Nagoya University, Nagoya, Japan in 1976, an M.S. degree in physics from Nagoya University in 1978, and a D.Sc. in physics from Nagoya University in 1981.
更新日期/Last Update:
2012-05-07