Previous [1] [2] [3] [4] [5] [6] [7] [8]

Journal of Inforamtion Science and Engineering, Vol.6 No.4, pp.325-337 (December 1990)
Environmental Studies of ICM Segmentation Algorithm*

Chaur-Chin Chen and Richard C. Dubes+
Department of Computer Science
National Tsing Hua University
Hsinchu 30043, Taiwan, R.O.C
+Computer Science Department
Michigan State University
East Lansing, Michigan 48824, U.S.A.

The ICM (Iterated Conditional Modes) algorithm has recently been exploited for image segmentation. The ICM segmentation algorithm iteratively updates the label of each pixel until a prescribed criterion is achieved. Most experiments and work have assumed that the true labeling is modeled by a discrete Markov random field and that the observed degraded image is formed by adding i.i.d. Gaussian noise to the true image. This paper reports on the ICM algorithm with various assumptions of degradation models. We characterize the mathematical formulas, list the ICM algorithm, and give the experiments based on known model parameters to segment synthetic images. The ICM algorithm segments images reasonably well under a variety of degradation models even though prior information is inadequate. A practical application of ICM algorithm for reconstructing an infrared image for target recognition is given.

Keywords: Gibbs distribution, ICM, Markov random field, pixel labeling, segmentation

Received December 15, 1989; revised July 15, 1990.
Communicated by Jun S. Huang.
*Dr. Chen was partially supported by NSC Grant 79-0408-E007-21.
Dr. Dubes was supported by NSF Grant IRI-8901513.
Part of this paper was presented in Taiwan-CVGIP workshop in 1989.