基于MRF模型的可靠的图像分割

本文提出一种可靠的图象分割算法。基于实际图象是分割图像叠加了不规则噪声的假设，用ＭＦＲ模型描述分割图象的先验分布，用被污染的高斯分布描述待分割的图像。采用Ｂａｙｅｓ方法，根据分割图像的后验分布所对应的ＭＲＦ模型的条件概率，用ＩＣＭ局部优化方法，获得ＭＡＰ准则下的图像分割结果。该算法与Ｌａｋｓｈｍａｎａｎ等提出的算法相比，具有更好的可靠性，实验结果是令人满意的。

Reliable Image Segmentation Based on Markov Random Field Model

A reliable image segmentation algorithm is proposed in this paper.Based on the assumption that the observed image is the sum of the segmentation image and the irregular corruptive noise,the segmentation image is modeled by a MRF(Markov random field)prior distribution while the observed image is modeled by a contaminated Gaussian distribution.The Bayes formulation is adopted to obtain the conditional distribution of the a posteriori distrbution of the segmentation image conditioned on the observed image,and based on MAP (maximum a posteriori) criterion the segmentation result is then obtained by applying ICM(iterated conditional mode) algorithm to maximize the a posteriori distribution.The algorithm has reliable segmentation result compared to that proposed by Lakshmanan et al .Experimental results are very satisfactory.