摘要
智能电表芯片图像中存在的亮度不均匀和字符大小不一等问题会影响二值化效果,导致字符识别率降低。而目前常用于处理亮度不均匀图像的局部阈值二值化算法大多都没有考虑图像的整体情况,容易产生过分割效应和伪影,且需要手动调整参数,无法有效处理大小不一的字符。针对上述问题,提出一种基于自适应混合阈值的二值化算法,改进了局部阈值的计算方法,利用笔画宽度变换自动计算窗口大小,并根据窗口大小计算自适应参数K,由此得到自适应的局部阈值,对局部和全局阈值进行加权,得到自适应的混合阈值。实验结果表明,相比传统的全局和局部阈值算法,该算法的评估指标和二值化效果都有一定的提升。
Uneven brightness and different character size in the chip image of smart meter will affect the binarization effect, which reduces the character recognition rate. However, most of the local threshold binarization algorithms commonly used to process images with uneven brightness do not consider the overall situation of the image, which is prone to over segmentation effect and artifacts. These algorithms need to manually adjust the parameters, and cannot effectively process characters of different sizes. To solve the above problems, we propose a binarization algorithm based on adaptive hybrid threshold. We improved the calculation method of local threshold, automatically calculated the window size by stroke width transformation, and calculated the adaptive parameter K according to the window size, so as to obtain the adaptive local threshold. And the local and global thresholds were weighted to get an adaptive hybrid threshold. The experimental results show that compared with the traditional global and local threshold algorithm, both the evaluation index and binarization effect of this algorithm are improved.
作者
刘朋远
田瑞
周媛奉
窦圣霞
樊博
胡循勇
Liu Pengyuan;Tian Rui;Zhou Yuanfeng;Dou Shengxia;Fan Bo;Hu Xunyong(State Grid Ningxia Marketing Service Center(Metrology Center),Yinchuan 750000,Ningxia,China;College of Electrical Engineering,Sichuan University,Chengdu 610065,Sichuan,China)
出处
《计算机应用与软件》
北大核心
2023年第1期210-215,共6页
Computer Applications and Software
关键词
智能电表
图像处理
二值化
笔画宽度变换
混合阈值
Smart meter
Image processing
Binarization
Stroke width transformation
Hybrid threshold
