基于互信息量的图像分割

图像分割是图像信息处理的热点和难点之一，常用的分割方法有阈值法和聚类法等．模糊C均值（FCM）算法因其实现简单、结果较优而得到广泛应用，但FCM算法存在过分依赖初值、收敛于局部极值和需预先给定分类类数等问题．研究者们对此进行了大量研究和改进，但均无法彻底解决上述问题，基于模拟退火算法和互信息量，以最大互信息量为优化目标，文中提出了一种新的分类类数判据一互信息熵差，并在此基础上构造了一种新的阈值分割算法——最大互信息量分割算法（MMS），实验结果表明，MMS克服了FCM算法的上述不足．更为重要的是，作为一种一般性的分类算法，MMS算法如同FCM一样，可以应用到图像分割以外的更广阔的领域，如经济学、运筹学、模式识别等．

Image Segmentation Based on Mutual Information

Image segmentation is of great importance in the field of image processing. A wide variety of approaches have been proposed for image segmentation. Among them, Fuzzy C-Means(FCM) is a classic one. FCM is a method of clustering which allows one piece of data to belong to two or more clusters. This method is frequently used in pattern recognition and image segmentation. But there are some disadvantages in FCM and many researches have been done to modify it. Mutual information (MI) is a basic concept from Shannon information theory and has been widely used in image registration before. Recently, some segmentation methods based on MI are reported. In this paper the authors propose a new segmentation method named MMS (Mutual information Maximized Segmentation), which is based on the optimization object of maximization of MI and a new measurement for determining the number of clusters in an image called dMI (difference of mutual information). In MMS method, simulated annealing is used for finding the global minimum. Experiments show that MMS has some virtues which FCM has not. It is more important that MMS can be used in other scope, such as economics, operational research and pattern recognition, just like FCM.