基于最大模糊指数熵的模糊目标分割算法  

Fuzzy Object Segmentation Based on Maximal Fuzzy Exponential Entropy

作　　者：刘锁兰[1] 杨静宇[1] 郭克华[1] 

机构地区：[1]南京理工大学计算机系,南京210094

出　　处：《计算机科学》2007年第9期240-241,272,共3页Computer Science

基　　金：国防基础研究基金(J1500C002)

摘　　要：针对红外目标主体模糊分割时易引起过或欠分割的问题,本文结合模糊集和熵理论提出了一种采用最大模糊指数熵准则的分割算法。根据像素灰度值将原始图像中像素分为黑和亮两个模糊集,利用模糊指数熵准则确定模糊区间的范围,再结合遗传算法强大的全局最优解搜索能力寻找模糊参数的最优组合,确定最佳阈值。对比实验表明,本文方法时主体模糊的目标分割效果明显优于最大类间距离方差法。但若向图像中添加大量随机噪声,两种方法都存在过分割现象。Because of the fuzzy infrared object, over-segmentation or less-segmentation is shown. Combining the theories of fuzzy sets and entropy, a new segmentation method of maximal fuzzy exponential entropy is proposed. The original image is segmented into dark and bright fuzzy sets based on pixel gray value. And based on the rule of fuzzy exponential entropy, the fuzzy scope is established. To look for the best segmentation threshold, genetic algorithm is used to search optimal combination. Contrast experiments show that segmentation results are better than OSTU＇S to fuzzy object. While rand noise is added, over segmentation is found in two methods.

关 键 词：模糊 熵理论 分割 遗传算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]