改进的基于二维直方图的最大模糊熵分割方法  

Improved Maximum Fuzzy Entropy Method Based on Two-Dimensional Histogram

作　　者：刘文波[1] 甄子洋[1] 刘媛媛[1] 张弓[2] 

机构地区：[1]南京航空航天大学自动化学院,南京210016 [2]南京航空航天大学信息科学与技术学院,南京210016

出　　处：《南京航空航天大学学报》2008年第5期692-695,共4页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：航空科学基金(04D52032)资助项目;江苏省自然科学基金(K2001047)资助项目

摘　　要：针对基于一维直方图的传统模糊熵算法对噪声敏感的问题,提出了一种新的基于二维直方图的最大模糊熵图像分割算法。新算法根据中心像素与邻域均值之间的关系划出二维灰度直方图中的有效区域,并考虑了图像的空间信息,将二维灰度直方图中像素的邻域均值的隶属度与中心像素隶属度相结合,得出新的隶属度计算方法,然后通过最大化模糊熵函数来确定图像的最优分割阈值。通过对实际图像的分割实验,表明了本文算法具有良好的去噪和图像细节保持能力。Aimed at the problem that traditional fuzzy entropy thresholding method based on one dimensional histogram is sensitive to the noise, a novel maximum fuzzy entropy thresholding method based on two-dimensional histogram is presented. The algorithm partitions off a validity range from the two-dimensional gray histogram based on the relation between the centre pixel and the neighborhood mean. Furthermore, to fuse the spatial information of the image, it redefines a new membership function and the function considers the influence of the neighborhood mean on the centre pixel. Therefore, the optimal thresholds are obtained by maximizing the new fuzzy entropy function. The segmentation results of the real images show that the proposed method has strong capabilities of removing noises and preserving image details.

关 键 词：图像分割 模糊熵 二维直方图 模糊隶属度函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]