Extraction of arbitrary text in natural scene image based on stroke width transform

Text extraction plays an important role in numerous applications. Research on its method still need to be improved in order to achieve better performance, to increase the reliability of text extraction system and to deal with complex cases of text extraction. The majority of the text extraction methods are focusing on horizontal and near horizontal text lines; however, text in natural scene might be in arbitrary line in real time. Thus, this paper aims to solve the issue of extracting the arbitrary oriented text by suggesting a method for text detection and localization based on the Stroke Width Transform. The proposed method is tested on an arbitrary text dataset and ICDAR dataset. The result of the experiment shows that the proposed method adapts well to the arbitrary text.