基于最小类内差和最大类间差的图像分割算法研究  

Image Segmentation Algorithm Research Based on Minimum Within-cluster Difference and Maximum Between-cluster Difference

作　　者：吴昊[1] 汪荣贵[2] 方帅[2] 杨万挺[2] 

机构地区：[1]合肥师范学院计算机科学与技术系,安徽合肥230061 [2]合肥工业大学计算机与信息学院,安徽合肥230009

出　　处：《工程图学学报》2011年第1期67-75,共9页Journal of Engineering Graphics

基　　金：国家自然科学基金资助项目(60705015);安徽省自然科学基金资助项目(070412054)

摘　　要：针对现有二维Otsu图像分割算法未考虑到目标和背景这二类像素自身的内聚性,提出一种新的自适应二维Otsu算法。该算法通过待分割图像的二维直方图,分别统计类内的绝对差、类间总体离差以反映类内、类间的离散度,从而构造出新阈值判别函数。通过一种改进的遗传算法优化二维阈值判别函数,自动得到较理想的分割阈值。实验结果表明,与其它阈值判别函数相比,通过优化新的阈值判别函数得到的二维阈值,具有了较好的分割效果,能够更好地保留了目标物的轮廓,而且计算量小。The present two-dimensional Otsu image segmentation algorithm ignores the cohesiveness of foreground and background pixels.A novel method of threshold recognition algorithm is proposed,which,by using two-dimensional histogram of the target image,counts the absolute difference within the cluster and the average total deviation between the cluster to reflect the scattered difference,and then constructs a new threshold recognition function.An improved genetic algorithm is adopted to optimize the new threshold recognition function so as to obtain the ideal threshold value automatically.Experimental results show that the two-dimensional threshold value obtained through the optimized threshold recognition function is of good segmentation efficiency,of good retaining of the object outline and of low work amount of calculation.

关 键 词：图像分割 二维OTSU法 阈值判别函数 类内最小差 类间最大差 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]