Text Localization and Recognition from Natural Scene Images using AI

In computer vision systems, text detection and recognition (TDR) in natural scene images can be used for things like license plate recognition, automated street sign interpretation, and assisting blind people. Accordingly, finding text within an image is a time-based challenge in the field of comput...

Full description

Saved in:
Bibliographic Details
Published in2022 International Conference on Automation, Computing and Renewable Systems (ICACRS) pp. 1153 - 1158
Main Authors Shekar Goud, D., Vigneshwari, M, Aparna, P, Vijayasekaran, G, Singh Yadav, Ajay, Kumar, Ashok
Format Conference Proceeding
LanguageEnglish
Published IEEE 13.12.2022
Subjects
Accuracy
Blindness
Computer vision
Deep Learning
Enhancement
Image recognition
Localization
Location awareness
Measurement
Natural images
Resize
Speech recognition
Text
Text recognition
Online AccessGet full text

Cover

More Information
Summary:In computer vision systems, text detection and recognition (TDR) in natural scene images can be used for things like license plate recognition, automated street sign interpretation, and assisting blind people. Accordingly, finding text within an image is a time-based challenge in the field of computer vision. Because of factors like cluttered backgrounds, image blurring, partially obscured text, various fonts, noise, and fluctuating lighting, text identification in natural scenes has become a significant task with the increase in the use of actual vision systems. Images and videos with accompanying textual data can be leveraged for automatic annotation. This study provides a system for automatically identifying the text from images, and it discusses the methodology behind locating and recognizing text in images of natural scenes. This article handles the scene text recognition challenge from start to finish, breaking it down into text localization and recognition. The Maximally Stable Extremal Regions (MSER) technique is used to detect text and non-text regions in images for localization purposes. Convolutional Neural Networks (CNNs) and convolutional recurrent neural networks (CRNNs) are utilized for text recognition. By evaluating the accuracy, precision, and F1 score, the CRNN is determined to be the best.
DOI:10.1109/ICACRS55517.2022.10029220