基于离线参数调整的古籍图像二值化算法  

Ancient document image binarization using offline parameter tuning

作　　者：冯炎 陈汝真 FENG Yan;CHEN Ru-zhen(School of Information Science and Technology,Tibet University,Lhasa 850000,China)

机构地区：[1]西藏大学信息科学技术学院,西藏拉萨850000

出　　处：《计算机工程与设计》2022年第3期794-799,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61661047)。

摘　　要：古籍退化对古籍图像二值化带来较大挑战,为此,提出一种基于离线参数调整的古籍图像二值化算法。算法分为两步,估计古籍图像背景局部均值,结合基于拉普拉斯能量的二值化算法对古籍进行二值化;根据迭代弗里德曼竞赛算法设计一种离线参数调整方法优化算法中的参数配置。所提算法在H-DIBCO 2016和H-DIBCO 2018数据集上的综合实验结果均优于其它二值化算法,验证了所提算法的有效性。There are many challenges addressed in ancient document image binarization due to the diversity of degradation effects,to deal with this,ancient document image binarization using offline parameter tuning was proposed,the algorithm was divided into two steps.The binarization algorithm was constructed based on background local mean estimation and Laplace energy-based binarization algorithm.An offline parameter tuning method based on the iterated Fried Man racing algorithms was studied to suggest the parameters for the proposed algorithm.The comprehensive experimental results on the H-DIBCO(Handwritten Document Image Binarization Contest)2016 and H-DIBCO 2018 show that the proposed algorithm outperforms other recent binarization techniques,the validity of the algorithm is indicated.

关 键 词：参数调整 古籍图像 二值化 图像修复 背景局部均值 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]