Abstract: The extraction of text in an image is a classical problem in the computer vision. With the increasing popularity of practical vision system text recognition in natural scenes becomes a critical task. Text data present in images and video contain useful information for automatic annotation, image indexing. But as natural scene images contain complex background, multiple fonts and orientations, and different alignments make the problem of automatic text extraction extremely challenging. This paper proposes the approach for automatic detection of text from images and explains the methodology to extract and recognize multi-oriented text in natural scene images.

Keywords: Text Localization, Text extraction, Text recognition, Sobel Mask, Binarization, Bounding Box.