Ticker text extraction from Bangla news videos

Abstract

In this paper, a framework for recognition of Bangla ticker text1 from the Bangla news videos is presented. Tesseract OCR [1] has been used for Bangla script recognition. Tesseract OCR gives good results for text recognition in documents. But in case of images and videos, some processing is required beforehand. Approach here is to provide processed images to the Tesseract OCR to get better results than directly providing the raw video frames to the Tesseract OCR. The ticker text recognized can further be used for indexing of news videos on the basis of recognized keywords. Indexing of news videos is important for news monitoring agencies. Till now this is done manually. Automation of the monitoring process and indexing the news videos can save a lot of time as well as the efficiency of the news monitoring system.