Abstract
Markov random field(MRF) models for segmentation of noisy images are discussed. According to the maximuma posteriori criterion a configuration of an image field is regarded as an optimal estimate of the original scene when its energy is minimized. However, the minimum energy configuration does not correspond to the scene on edges of a given image, which results in errors of segmentation. Improvements of the model are made and a relaxation algorithm based on the improved model is presented using the edge information obtained by a coarse-to-fine procedure. Some examples are presented to illustrate the applicability of the algorithm to segmentation of noisy images.
Similar content being viewed by others
References
S. Geman, D. Geman,IEEE Trans. on Pat. An. Mach. Intell.,PAMI-6(1984)6, 721–741.
J. Besag,J. Roy. Statist. Soc. B,48(1986)3, 259–302.
Kuang Jinyu, Yao Xiaoyan,Chinese Journal of Computers 14(1991)7, 514–522, (in Chinese)
H. Derin, H. Elliott,IEEE Trans. on Pat. An. Mach. Intell,PAMI-9(1987)1, 39–55.
R. Hu, M. M. Fahmy,Signal Processing,26(1992)3, 285–305.
C. S. Won, H. Derin,CVGIP: Graphical Models and Image Processing,54(1992)4, 308–328.
Guo Yuchun, Kuang Jinyu,Journal of Beijing Normal University (Natural Science),30(1994)1, 60–66, (in Chinese).
D. Geman et al.,IEEE Trans. on Pat. An. Mach. Intell.,12(1990)7, 609–628.
R. Kindermann, J. L. Snell, Markov Random Fields and Their Applications, Providence RI: Amer. Math. Soc., 1980.
D. Marr, E. Hildreth,Proc. Royal Soc. London B,207(1980), 187–217.
A. Kundu,Pattern Recognition,23(1990)5, 423–440.
S. Mallat, W. L. Hwang,IEEE Trans. on Inform. Theory,IT-38(1992)2, 617–643.
Author information
Authors and Affiliations
Additional information
Supported by the National Natural Science Foundation of China.
An erratum to this article is available at http://dx.doi.org/10.1007/BF02685839.
About this article
Cite this article
Jinyu, K., Junxiu, Z. On Markov random field models for segmentation of noisy images. J. of Electron.(China) 13, 31–39 (1996). https://doi.org/10.1007/BF02684712
Issue Date:
DOI: https://doi.org/10.1007/BF02684712