Paper Info
Title
Oriental-script text detection and extraction in videos
Abstract
Urdu text detection and extraction from videos especially from news feed is a challenging task. The main challenge is to design a robust algorithm that can accurately detect and localize Urdu text regions in videos. Urdu language is originated from oriental languages like Arabic, Persian and Turkish, so alphabets of these languages are very similar. Due to the similarities of basic alphabets and writing styles, this research is equally applicable to Arabic or Persian text. The development of an OCR that gives acceptable results on Urdu is still an ongoing research area. Thus need of a good text detection and extraction system is needed that can reduce the recognition errors as much as possible. It is because if the input image contains only text candidate regions, then possibilities of errors will be reduced for the same OCR. The proposed system is evaluated on a set of videos of news channels that vary along the dimensions of intensity variations, scale and orientation. Accuracy in terms of precision and recall rates is analyzed to determine the success and limitations of the approach. Problems, possible errors and suggestions for improvements are also discussed in the paper.
Year
DOI
Venue
2017
10.1109/ASAR.2017.8067752
2017 1st International Workshop on Arabic Script Analysis and Recognition (ASAR)
Keywords
Field
DocType
Urdu Text Detection,Urdu Text Localiztion,Projection Profiles,Optical Character Recognition,Content Based Image Retrieval
Turkish,Algorithm design,Arabic,Computer science,Persian,Writing style,Precision and recall,Speech recognition,Urdu,Natural language processing,Artificial intelligence,Text detection
Conference
ISBN
Citations 
PageRank 
978-1-5090-6629-2
1
0.37
References 
Authors
0
2
Name
Order
Citations
PageRank
Usman Shahzad110.37
Khurram Khurshid212915.94