A Review on Arbitrarily and Horizontal Scene Text Extraction from Images or Videos

Abstract

The extraction of text in an image is classical problem in the computer vision. With the increasing popularity of practical vision system text recognition in videos becomes a critical task. Text data present in images and video contain useful information for automatic annotation, image indexing. But as video frames contain complex background, multiple fonts and orientation, and different alignments make the problem of automatic text extraction extremely challenging. There are two types of text in video frames. One is scene text and other is a graphic text. Scene text is captured in scene while recording while graphic text is more structured and closely related to subject. Scene text is difficult to detect than a graphic text. In proposed technology extraction of scene text is done.