二维最大熵和二维最小交叉熵结合的图像分割  

Image Segmentation Based on 2-D Maximum Entropy and 2-D Minimum Cross Entropy

作　　者：张新明[1] 孙全党[1] 郑延斌[1] 张慧云[1] 李双[1] 

机构地区：[1]河南师范大学计算机与信息技术学院,河南新乡453007

出　　处：《电光与控制》2011年第2期54-59,共6页Electronics Optics & Control

基　　金：河南省重点科技攻关项目(092102210017;102102210180);河南省教育厅科技攻关项目(2008B520021)

摘　　要：二维最大熵法和二维最小交叉熵法是目前常用的两种阈值分割方法,但在某些时候因为两种方法获取的阈值过高或者过低,使得分割失效。针对此问题,提出了基于二维最大熵法和二维最小交叉熵法结合的图像分割方法。首先,对二维最小交叉熵公式进行转化;然后,利用多目标规划理论将这两种方法有机结合使得到的阈值既满足二维最大熵原则,又满足二维最小交叉熵原则;最后,利用二维直方图的特点推导出新型递推算法搜索最佳阈值并降低计算复杂度。仿真实验结果表明,本文提出方法不仅有效,弥补了两者在某些应用上都不能有效分割的不足,而且分割时间少,约为0.3 s。The thresholding method based 2-D maximum entropy and the one based on 2-D minimum cross entropy are used widely in image segmentation today, but in some applications, they fail to segment images because of too high or too low thresholds. Therefore, we proposed an image thresholding method based on the combination of 2-D maximum entropy and 2-D minimum cross entropy. Firstly, the ＇formula of the 2-D minimum cross entropy was transformed, then 2-D maximum entropy and 2-D minimum cross entropy were combined together using multi-objective programming theory so that the optimal threshold value could satisfy the threshold requirement of the both. A new recursive algorithm was inferred using the features of the 2-D histogram in order to search the best threshold vector and to reduce the computing complexity. Experimental results show that： 1 ） the proposed method is effective and can compensate the shortcomings of the two methods in some applications ; and 2） the computation time is less, which is about 0.3 second.

关 键 词：图像分割 二维最小交叉熵 二维最大熵 递推算法 多目标规划 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP391.41[自动化与计算机技术—计算机应用技术]