摘要
针对中文版面扫描后生成图像存在的倾斜问题,提出了一种自适应中文版面倾斜校正方法,首先设计一种基于布局特征的图像分类算法,从而充分挖掘文本数据之间的内在联系,快速区分不同类型文本,接着根据区分的文本类型自适应采取不同策略估计倾斜角度,从而最终实现对中文版本扫描图像的倾斜校正.实验结果表明,与k近邻聚类、傅立叶变换、霍夫变换和投影轮廓分析等算法相比,所提方法综合指标性能最优.
A self-adaptive Chinese layout tilt correction method was proposed to address the issue of tilt in images generated after scanning Chinese layouts.Firstly,an image classification algorithm based on layout features was designed to fully explore the inherent connections between text data and quickly distinguish different types of text.Then,different strategies are adopted to adaptively estimate the tilt angle based on the distinguished text types,ultimately achieving tilt correction of Chinese version scanned images.In the experimental stage,compared with algorithms such as k-nearest neighbor clustering,Fourier transform,Hough transform,and projection contour analysis,the proposed method had shown the best comprehensive performance indicators.
作者
郭丹
GUO Dan(Science and Technology Department,Xi′an Medical University,Xi′an 710021,China)
出处
《云南师范大学学报（自然科学版）》
2024年第6期48-51,共4页
Journal of Yunnan Normal University:Natural Sciences Edition
基金
陕西省哲学社会科学研究专项青年资助项目(2023QN0277).
关键词
扫描图像
倾斜校正
图像分类
自适应
Scanning images
Tilt correction
Image classification
Self-adaption
