基于分解的二维指数灰度熵图像阈值分割  

Two-dimensional exponential gray entropy image thresholding based on decomposition

作　　者：吴诗婳[1] 张晓杰[1] 吴一全[1,2] 

机构地区：[1]南京航空航天大学信息科学与技术学院,南京210016 [2]南京大学计算机软件新技术国家重点实验室,南京210093

出　　处：《中国体视学与图像分析》2010年第4期351-358,共8页Chinese Journal of Stereology and Image Analysis

基　　金：国家自然科学基金项目(60872065);南京大学计算机软件新技术国家重点实验室开放基金项目(KFKT2010B17)

摘　　要：现有最大Shannon熵或Tsallis熵阈值选取方法仅仅依赖于图像灰度直方图的概率信息,而没有直接考虑类内灰度的均匀性;且用对数定义的Shannon熵还存在无定义值和零值的问题。为此,提出了一维和基于分解的二维指数灰度熵阈值分割方法。首先给出了指数灰度熵的定义及其一维阈值选取方法;然后将其推广得到二维指数灰度熵阈值选取公式,提出了基于分解的二维指数灰度熵阈值分割方法。通过分别求原像素灰度级图像和邻域平均灰度级图像的一维指数灰度熵最佳阈值,并将其组合求解二维指数灰度熵最佳阈值,大大缩小了搜索空间,使计算复杂度由O(L4)降为O(L)。实验结果表明,与基于基本粒子群算法的二维最大Shannon熵法、二维最大Tsallis熵法以及采用递推的二维交叉熵法相比,所提出方法的分割效果具有明显的优势,且运行时间大幅减少。The method of threshold selection based on two-dimensional maximal Shannon or Tsallis entropy only depends on the probability information from gray histogram of image,and does not directly consider the uniformity of within-cluster gray level.The Shannon entropy defined by logarithm has the problem of undefined value and zero value.Thus,a one-dimensional and two-dimensional exponential gray entropy thresholding method is proposed.Firstly,a definition of the exponential gray entropy is given,and one-dimensional exponential gray entropy method for threshold selection is derived.Then,it is extended and two-dimensional exponential gray entropy thresholding method based on decomposition is proposed.The optimal threshold of one-dimensional exponential gray entropy method for pixel gray-level image or neighborhood average gray-level image is computed,respectively,and they are combined to obtain the optimal threshold of the two-dimensional exponential gray entropy method.As a result,the search space is significantly reduced.The computation complexity is reduced from O（L4） to O（L）.The experimental results show that,compared with two-dimensional maximal Shannon or Tsallis entropy thresholding method based on particle swarm optimization（PSO） and two-dimensional cross entropy method with recursion,the method proposed in this paper has much superior segmentation performance and its running time is reduced significantly.

关 键 词：图像分割 阈值选取 二维指数灰度熵 分解 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]