Abstract—Daily lives contain many of the menus and
signboards which carry important information, but sometimes it
may cause problems when we cannot understand it. This paper
aimed to develop a new system that translates Arabic texts of the
signboards into English text by using mobile phone camera. In
spite of the diversity of translation products of text embedded in
images for many languages, Arabic texts seem to be not yet well
solved to address this problem. The system will automatically
translate the Arabic text embedded in images into English
language. Four subsystems used in the algorithm: preprocessing,
segmentation (text detection), character recognition and
translation. Dealing with Arabic language was the most
important problem faced by the proposed system because it has
a set of characteristics makes the identification very difficult,
such as words interrelated. The system automatically detects the
text and works well with different backgrounds, rotated images,
skewed, font sizes and blurred images. The system was assessed
using recall measurement to evaluate the performance of the
developed system and the experimental results of character
recognition show a rate of 81.82%, the word recognition
subsystem gave a rate of (94.44%) and the word translation was
about (83.33%).
Index Terms—Arabic translation, information extraction,
character recognition.
The authors are with Al-Hussein Bin Talal University (AHU), Jordan
(e-mail: Rafiq_alhashimy@yahoo.com).
Cite:Rafeeq Abdul Rahman A. Al-Hashemi and Shoroq Almamon Alsharari, "Instant Arabic Translation System for Signboard Images Based on Printed Character Recognition," International Journal of Machine Learning and Computing vol.3, no. 4, pp. 384-388, 2013.