DocumentCode :
2611854
Title :
MRF model and edge-preserving image restoration with neural network
Author :
Xiaoyu, Qiao ; Enfang, Sang ; Bedini, Luigi ; Tonazzini, Anna
Author_Institution :
Underwater Acoust. Dept., Harbin Eng. Univ., China
Volume :
2
fYear :
1997
fDate :
28-31 Oct 1997
Firstpage :
1432
Abstract :
In the Bayesian approach, using Markov random field (MRF) models, prior knowledge is incorporated into the problem via a parametric Gibbs prior, and the solution is obtained by minimizing the resulting posterior energy function. There are two major difficulties with this approach: the non-convexity of the function to be optimized and the choice of the MRF model parameters that best fit the available prior knowledge. Since these parameters affect the quality of the reconstruction considerably, selecting them is a very critical task. This paper deals with the restoration of piecewise smooth images. A trained generalized Boltzmann machine can then be used in connection with a Hopfield analogue circuit to implement a mixed-annealing scheme for the minimization of the non-convex posterior energy
Keywords :
Bayes methods; Boltzmann machines; Hopfield neural nets; Markov processes; analogue processing circuits; concave programming; image restoration; minimisation; parameter estimation; simulated annealing; Bayesian approach; Hopfield analogue circuit; Hopfield neural net; Markov random field models; edge-preserving image restoration; image reconstruction quality; mixed-annealing scheme; nonconvex function optimization; parameter selection; parametric Gibbs prior; piecewise smooth images; posterior energy function minimization; prior knowledge; trained generalized Boltzmann machine; Acoustical engineering; Bayesian methods; Circuits; Degradation; Image reconstruction; Image restoration; Knowledge engineering; Markov random fields; Minimization; Neural networks; Power engineering and energy;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Intelligent Processing Systems, 1997. ICIPS '97. 1997 IEEE International Conference on
Conference_Location :
Beijing
Print_ISBN :
0-7803-4253-4
Type :
conf
DOI :
10.1109/ICIPS.1997.669254
Filename :
669254
Link To Document :
بازگشت