A segmentation-free method for image retrieval and pattern spotting in historical documents using convolutional features

This paper describes a new segmentation-free method for retrieving images and spotting patterns in historical document image collections. The proposed method needs no training on the target domain, characterizing a problem-independent approach. For this purpose, the query and the document image repr...

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications Vol. 84; no. 32; pp. 39635 - 39665
Main Authors: Curi, Zacarias, Nicolas, Stéphane, Tranouez, Pierrick, M. Saavedra, Jose, de Souza Britto Jr, Alceu, Heutte, Laurent
Format: Journal Article
Language:English
Published: New York Springer US 01.09.2025
Springer Nature B.V
Springer Verlag
Subjects:
1250: Heritage Preservation in the Digital Age: Advances in Machine Learning
Artificial Intelligence
Computer Communication Networks
Computer Science
Computer Vision and Pattern Recognition
Cross correlation
Data Structures and Information Theory
Datasets
Deep learning
Digitization
Document and Text Processing
Documents
Feature maps
Human-Machine Interaction
Image retrieval
Image segmentation
Logos
Metadata
Monomodal and Multimodal Processing
Multimedia Information Systems
Queries
Retrieval
Special Purpose and Application-Based Systems
Document retrieval
Pattern spotting
Fully-convolutional activation features
Cross-correlation
Historical documents
Fully-convolutional Networks
ISSN:1573-7721, 1380-7501, 1573-7721
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper describes a new segmentation-free method for retrieving images and spotting patterns in historical document image collections. The proposed method needs no training on the target domain, characterizing a problem-independent approach. For this purpose, the query and the document image represented by feature maps extracted using intermediate layers of a pre-trained Fully Convolutional Network are submitted to a cross-correlation process. The produced similarity heatmap is used to locate the query occurrences on the document page. A robust experimental protocol using three datasets shows promising results on image retrieval and pattern spotting. The experiments conducted on the public DocExplore dataset demonstrated that the proposed method could improve the mAP by 69.1% for the spotting task and 21% for the image retrieval task compared to the state-of-the-art. Additional experiments on image retrieval using the Horae dataset and logo spotting using the Tobacco800 dataset confirmed the high generalization capability of the proposed method.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1573-7721
1380-7501
1573-7721
DOI:10.1007/s11042-025-20766-6