灰度图像的二维交叉熵阈值分割法投出

__CROBUTTON Ac__ptAllChangesInDoc 灰度图像的二维交叉熵阈值分割法范九伦 = 1 \* GB3 ① 雷博 = 1 \* GB3 ①, = 2 \* GB3 ②(

1 灰度图像的二维交叉熵阈值分割法 ①①② , 范九伦雷博 (①.,710061; 西安邮电学院信息与控制系陕西西安 ②.,710071) 西安电子科技大学电子工程学院陕西西安 摘要 OtsuLiLee :一维法是一个经典的阈值分割方法,遵循该方法的构造思想,与基于交叉熵提出了一 LiLee__ 个阈值分割方法。本文在解释和说明与的方法的基础上,将其到二维灰度直方图上,提出了二 Otsu 维交叉熵的图像分割算法并给出快速递推公式。与二维法相比,本文方法能够更好的适应目标和背景 方差相差较大的情形,是一个有效的阈值分割方法。 关键词 Otsu :阈值分割;法;交叉熵 中图分类号:文献标识码: TN911.73A Two-dimensionalcross-entropythresholdingsegmentation methodforgray-leveli__ges ①①② , (FANJiu-lunLEIBo) (DepartmentofInfor__tionandControl,Xi’anInstituteofPostandTelecommunications,,Xi’an,Shaanxi710061,China; ①. .SchoolofElectronicEngineeringofXidianUniversity,Xi’anShaanxi710071,China) ② Abstract: One-dimensionalOtsu’smethodisaclassicalthresholdingsegmentationmethod.Li&Lee proposedathresholdingsegmentationmethodbasedoncross-entropyfollowingthisthought.Inthispaper,we extendthecross-entropymethodtotwodimensionsandpresentafastrecursiveformulabasedontheexplanation andillustrationofLi&Lee’smethod.Comparedwiththetraditionaltwo-dimensionalOtsu’smethod,thenew methodcanbebetteradaptedtothecasesthatthevarian__betweentheo__ectandthebackgroundislarge.Itis validatedthatthenewmethodisan__ailablethresholdingselectionmethod. Keywords: thresholdsegmentation;Otsu’smethod;cross-entropy 1 引言 图像分割是图像分析、理解和计算机视觉中的难点。在图像分割的诸多方法中,阈值化 [1][2] OtsuOtsu 技术是一种简单有效的方法。法是广泛使用的阈值分割方法之一,法也称为最 [3][4] c-Kurita 大类间方差法或最小类内方差法,等效于一维硬均值聚类算法。等在各类方差 Otsu[3][4] 相等的约束下运用条件相关混合概率模型对法进行了解释,从文和的描述可见 Otsu 法在理论上适用于目标和背景方差相差不大的混合正态分布情形。 Otsu 法涉及到阈值、目标均值和背景均值,如果用与构造的 “”Otsu“” 二值图像作为待分割图像的理想图像,则法的基本思想是从待分割图像和理想图像 1 60572133 国家自然科学基金资助(编号:)

腾讯文库灰度图像的二维交叉熵阈值分割法投出