基于模糊熵的自适应多阈值图像分割方法  

An Adaptive Multi-threshold Image Segmentation Method Based on Fuzzy Entropy

作　　者：宋欢欢[1] 李雷[2] 

机构地区：[1]南京邮电大学理学院应用数学研究中心,江苏南京210023 [2]南京邮电大学理学院,江苏南京210023

出　　处：《计算机技术与发展》2014年第12期32-36,共5页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61070234;61071167)

摘　　要：模糊技术能够很好地表达和处理不确定问题,是图像处理领域中一种非常重要的技术。基于模糊理论,文中提出了基于模糊熵的自适应多阈值分割方法。根据图像像素的概率分布,将图像行区域化,利用每个区域中像素属于前景和背景的模糊性定义隶属度函数,采用一维搜索方法确定最佳的隶属度函数窗宽,计算最大模糊熵,从而找到区域最优阈值。文中对多目标、光照不均匀、存在噪声和分割不完全的图像进行实验,结果表明该方法能够很好地解决上述问题,并且较传统的基于Otsu和模糊熵的图像单阈值分割方法,效果显著提高,具有较好的适应性和实用性。Fuzzy technology,which can well express and deal with uncertain problems,is very important and useful in the field of image processing. Based on the fuzzy theory,an Adaptive M ulti- threshold M ethod（ AM M- FE） of image segmentation based on fuzzy entropy is proposed. According to the distribution probability of the image pixels,divide the image into a plurality of regions. Define the membership functions using the pixels belonging to the foreground and the background blur in each area,determine the windowwidth of fuzzy membership function by using one- dimension search method,calculating the maximum fuzzy entropy to get the regional optimal threshold. By using multi- objective,non- uniform illumination,presence of noise and imperfect image in the experiment,the results showthat this method can greatly overcome these incomplete segmentation situations. Compared with traditional single threshold image segmentation methods like Otsu and fuzzy entropy,the effect of this method is significantly improved,which indicates that the proposed AM M-FE method has better adaptability and practicality.

关 键 词：图像分割 模糊熵 隶属度函数窗宽 多阈值图像分割 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]