摘要
文章提出一种基于图像处理的文档图像矫正系统,该系统首先对文档图像进行高斯滤波、局部边缘提取及膨胀操作,然后提取轮廓,最后基于Hough变换算法找到可以覆盖所有轮廓的最小矩形并生成仿射变换矩阵进行变换矫正。该方法相比传统的基于Hough变换的方法,增加基于图像处理的滤波及形态学处理,使矫正效果更好;相比其他常见的基于模式识别进行图像矫正的方法,抗干扰能力更强;相比常见的基于深度学习进行图像矫正的方法,运算速度快,模型小,更适合在智能手机等移动端使用。
We propose a document image rectification system based on image processing.The system first performs Gaussian filtering,local edge extraction,and dilation on the document image.Then,the system captures the contours and finally finds the smallest rectangle that can cover all contours based on the Hough transform algorithm,and generates an affine transform matrix for rectification.Compared to the traditional Hough transform-based method,our method first performs image process-based filtering and morphological processing on the document image,which can further improve the rectification performance.Compared to the traditional image rectification methods based on pattern recognition,our method has the better anti-interference ability.Compared to the image rectification methods based on Deep Learning,our models are faster and smaller,which is more suitable for mobile devices such as smartphones.
作者
孙丹阳
邵坤艳
孙春志
程文静
舒方杰
SUN Dan-yang;SHAO Kun-yan;SUN Chun-zhi;CHENG Wen-jing;SHU Fang-jie(Henan Province Engineering Research Center of Microcavity and Photoelectric Intelligent Sensing,School of Electronics and Electrical Engineering,Shangqiu Normal University,Shangqiu 476000,China)
出处
《电脑与信息技术》
2023年第3期20-23,共4页
Computer and Information Technology
基金
国家自然科学基金项目(项目编号:12004238)
商丘师范学院教育教学改革项目(项目编号:2021jgzdxm03)。
