Abstract:Document Layout Analysis (DLA) is critical for understanding and reconstructing documents, which aims to automatically recognize the layout structure of basic or semantic elements within a document. However, a DLA process faces certain challenges due to the diversity and complexity of document layouts with a variety of languages. In particular, it needs more theoretical and practical investigations for analyzing Chinese documents. This paper proposes a fusion framework of Whitespace Smear Cutting (WSC) and Swin Transformer for layout analysis, mainly in Chinese documents. Specifically, in the first phase, we perform a new kind of unsupervised segmentation of document images with our proposed WSC algorithm that can preserve the delicate edges of the connected blocks of a document. In the second phase, we utilize a novel semantic segmentation network based on the Swin Transformer for pixel-level classification. We design a new training paradigm of continuous training for the Swin Transformer, which consists of pre-training on the large-scale data and fine-tuning on the specific datasets to adapt the model to the special data distributions. In our fusion process, we utilize the pixel-level semantic information to direct and integrate the same semantic connected blocks obtained from the WSC algorithm and semantic segmentation with certain rules based on the confidence levels and block distributions, which effectively alleviates the challenging problem of bounding box overlap and thus improves the accuracy of semantic classification. Finally, it is demonstrated by the experimental results on a collected dataset of Chinese documents and the POD dataset that our proposed fusion framework is feasible and effective on DLA.

A Connected Components Based Layout Analysis Approach for Educational Documents

CNN Based Page Object Detection in Document Images

A New Component Based Algorithm for Newspaper Layout Analysis

Unified HMM-based Layout Analysis Framework and Algorithm

A Hybrid Approach for Document Layout Analysis in Document images

WeLayout: WeChat Layout Analysis System for the ICDAR 2023 Competition on Robust Layout Segmentation in Corporate Documents

A Handwriting Textline Extraction Approach Based on Connected Domain

Doc-GCN: Heterogeneous Graph Convolutional Networks for Document Layout Analysis

Document Layout Analysis and Reading Order Determination for a Reading Robot

Document AI: A Comparative Study of Transformer-Based, Graph-Based Models, and Convolutional Neural Networks For Document Layout Analysis

Graph-based Layout Analysis for PDF Documents

Integration of Text Information and Graphic Composite for PDF Document Analysis

VSR: A Unified Framework for Document Layout Analysis Combining Vision, Semantics and Relations

An Efficient Pre-Processing Method To Identify Logical Components From Pdf Documents

A Fusion Framework of Whitespace Smear Cutting and Swin Transformer for Document Layout Analysis

Pattern Analysis and Digitization Modeling of Papercutting Textures

Graphic Composite Segmentation for PDF Documents with Complex Layouts

Beyond document object detection: instance-level segmentation of complex layouts

Layout Analysis Algorithm Based on Probabilistic Graphical Model for Dunhuang Historical Documents

Image Layer Modeling for Complex Document Layout Generation.

Cross-Domain Document Layout Analysis Using Document Style Guide