摘要
利用大津法和迭代法进行汉字图像处理时,阈值选取是关键。本文结合实验仿真和直方图对文字图像进行识别,当图片尺寸较大或图像像素较高时,大津算法需要进行较高次数循环以便确定类间方差最大的阈值。迭代算法少量几次就可以找到满足条件的阈值,且处理的文字边缘平滑度、文字粘连性和清晰度效果优于大津法,有利于后续文字分割。
Threshold selection is the key to Chinese character image processing based on Otsu method and improved iterative method.Combined with experimental simulation and histogram for character image recognition,when the actual size of the image is large or the image pixels are high,Otsu algorithm needs to execute higher cycles each time to determine the threshold to maximize the inter class variance.The improved iterative algorithm can find the threshold that meets the conditions a few times,and the effect of text edge smoothness,text adhesion and clarity are better than Otsu method,which is conducive to subsequent text segmentation.
作者
马海云
张敬花
MA Haiyun;ZHANG Jinghua(School of Electronic Information and Eletrial Engineering,Tianshui Normal University,Tianshui 741001,China;School of Marxism,Tianshui Normal University,Tianshui 741001,China)
出处
《中央民族大学学报（自然科学版）》
2022年第3期76-80,共5页
Journal of Minzu University of China(Natural Sciences Edition)
基金
甘肃省科技计划项目(21JR7RE147)
天水师范学院教研教改项目(JY203056)
2021天水师范学院校级课程思政示范项目(SFXM2021014)。
关键词
古典文献文字识别
文字图像
阈值
改进迭代法
大津法
classical literature character recognition
character image
threshold
improved iterative method
otsu method
