Multilingual Text Detection and Identification from Indian Signage Boards
R. Rahul, Sreebha Bhaskaran, J. Amudha, Deepa Gupta · 2018
Text detection is an important and challenging field of research because of various complexities like different sizes and fonts of text, line orientation, different illumination conditions and complex backgrounds in natural scene images. In recent years, there has been significant progress in detection and identification of text from scene images. However, existing approaches haven't still been able to entirely overcome these challenges. Also, there hasn't been much work for Indian languages with respect to natural scene images. India is a land rich in languages and many languages has similar scripts, hence it is a challenge to differentiate between them. Also in India, in some cases, natural scene images will have multiple languages, the local language, Hindi and English. The focus of this paper is to detect and identify text belonging to Kannada, Hindi and English. In this paper, we propose to use Stroke Width Transform (SWT) algorithm for text detection. A language identification technique using tree bagging algorithm is also integrated along with SWT algorithm.