一种改进的Otsu阈值分割算法  

An improved Otsu threshold segmentation algorithm

作　　者：周迪[1] 夏哲雷[1] 

机构地区：[1]中国计量大学信息工程学院,浙江杭州310018

出　　处：《中国计量学院学报》2016年第3期319-323,344,共6页Journal of China Jiliang University

基　　金：浙江省自然科学基金资助项目(No.LY12F01011)

摘　　要：当目标与背景的类内方差差别较大时,Otsu法分割阈值偏向于方差较大的一类,从而使分割的效果不佳.针对这一问题,提出了一种改进Otsu阈值分割方法.利用目标在图像中所占比例,综合类内方差与类间方差,修正最佳阈值选择公式,并缩小算法遍历灰度范围来提高算法运算速率,最后通过区域一致性对算法的分割效果进行评价.实验结果表明,改进的Otsu法分割效果较好,并提高了算法执行效率.When the difference of intraclass variance between the target and the background is significant,the traditional Otsu threshold segmentation method will be biased toward the class variance within a large category.To solve the problem,an improved method was proposed.The proportion of the target in the image,combined with intraclass variance and interclass variance,was used to fix the best threshold selection formula and to improve the operation rate of the algorithm.The segmentation algorithm was evaluated by regional consistency.The experimental results show that the improved Otsu segmentation method can improve the efficiency.

关 键 词：OTSU法 阈值分割 类间方差 类内方差 区域一致性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]