Scene text image correction based on transformer

Ming Jin; Xiaosheng Tu; Jinfeng Han; Congyan Chen

doi:10.1117/12.2685812

15 August 2023 Scene text image correction based on transformer

Ming Jin, Xiaosheng Tu, Jinfeng Han, Congyan Chen

Proceedings Volume 12719, Second International Conference on Electronic Information Technology (EIT 2023); 127191Q (2023) https://doi.org/10.1117/12.2685812
Event: Second International Conference on Electronic Information Technology (EIT 2023), 2023, Wuhan, China

Abstract

In this work, we propose a new algorithm framework based on Transformer to address the distortion problem of scene text images caused by shooting angles or lighting conditions. This new algorithm framework is called Scene text image transformer (StTr). Specifically, we use a feature extractor containing six convolution modules for feature extraction. Then, we flatten the obtained feature maps and input them into the Transformer encoder and decoder. After obtaining the displacement matrix, we use it to perform geometric correction on the input image. We also use Transformer to eliminate the lighting problem and improve the accuracy of OCR recognition. We evaluated our StTr on different datasets and achieved a 16.73% improvement in accuracy compared to the state-of-the-art methods. Additionally, our model has the advantages of being lightweight and efficient.

Citation Download Citation

Ming Jin, Xiaosheng Tu, Jinfeng Han, and Congyan Chen "Scene text image correction based on transformer", Proc. SPIE 12719, Second International Conference on Electronic Information Technology (EIT 2023), 127191Q (15 August 2023); https://doi.org/10.1117/12.2685812

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available

Members: $17.00

Non-members: $21.00 ADD TO CART

PROCEEDINGS
8 PAGES

DOWNLOAD PAPER SAVE TO MY LIBRARY

GET CITATION

RIGHTS & PERMISSIONS

Get copyright permission Get copyright permission on Copyright Marketplace

KEYWORDS

Transformers

Light sources and illumination

Optical character recognition

Performance modeling

Education and training

Data modeling

3D modeling

Show All Keywords

Keywords/Phrases

Search In:

Publication Years