CASIA OpenIR  > 模式识别国家重点实验室  > 模式分析与学习
Page Segmentation for Historical Handwritten Documents Using Fully Convolutional Networks
Xu Y(徐玥)1,2; He WH(何文浩)1,2; Yin F(殷飞)2; Liu CL(刘成林)1,2
2017
Conference Name2017 14th IAPR International Conference on Document Analysis and Recognition
Pages541~546
Conference Date2017.11.9 - 2017.11.15
Conference PlaceKyoto, Japan
Abstract
Page segmentation is a fundamental and challenging
task in document image analysis due to the layout diversity.
In this work, we propose a pixel-wise segmentation method
for historical handwritten documents using fully convolutional
network (FCN). The document image is segmented into different
regions by classifying pixels into different categories:
background, main text body, comments, and decorations. By
supervised learning on document images with pixel-wise labels,
the FCN can extract discriminative features and perform pixelwise
segmentation accurately. After pixel-wise classification, postprocessing
steps are taken to reduce noises, correct wrong
segmentations and find out overlapping regions. Experimental
results on the public dataset DIVA-HisDB containing challenging
medieval manuscripts demonstrate the effectiveness and superiority
of the proposed method, which yields pixel-level accuracy
of above 99%.
Other Abstract版面分割是文档图像分析的基础任务,由于文档版面的多样,版面分割成为非常具有挑战性的任务。在这篇文章中,我们针对古籍手写文档,提出了一种基于全卷积神经网络的像素级的分割算法。通过对像素点进行分类,文档图像被划分为几个不同的区域:背景、正文、旁注以及装饰。通过对文档图像中像素类别的学习,全卷积神经网络可以提取出具有判别性的特征,并生成高准确率的像素级分割结果。在进行像素级的分类后,会进行后处理以减少噪声、更正错误的分割并找出重叠区域。公开数据集DIVA-HisDB包含多种具有挑战性的中世纪手写文档,在这一数据集上我们的算法实验取得了超过99%的准确率,表明了我们算法的准确度和优越性。
KeywordPage Segmentation Layout Analysis Fully Convolutional Network
Language英语
Document Type会议论文
Identifierhttp://ir.ia.ac.cn/handle/173211/20019
Collection模式识别国家重点实验室_模式分析与学习
Affiliation1.中国科学院大学
2.中国科学院自动化研究所,模式识别国家重点实验室
Recommended Citation
GB/T 7714
Xu Y,He WH,Yin F,et al. Page Segmentation for Historical Handwritten Documents Using Fully Convolutional Networks[C],2017:541~546.
Files in This Item: Download All
File Name/Size DocType Version Access License
Page Segmentation fo(1424KB)会议论文 开放获取CC BY-NC-SAView Download
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[Xu Y(徐玥)]'s Articles
[He WH(何文浩)]'s Articles
[Yin F(殷飞)]'s Articles
Baidu academic
Similar articles in Baidu academic
[Xu Y(徐玥)]'s Articles
[He WH(何文浩)]'s Articles
[Yin F(殷飞)]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[Xu Y(徐玥)]'s Articles
[He WH(何文浩)]'s Articles
[Yin F(殷飞)]'s Articles
Terms of Use
No data!
Social Bookmark/Share
File name: Page Segmentation for Historical HandwrittenDocuments Using Fully Convolutional Networks.pdf
Format: Adobe PDF
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.