Text extraction from natural images based on stroke width map
Chunmei Liu · 2013
This paper proposes a text extraction method for text in natural images, which apply color and edge information to obtain text candidate areas, and use stroke width information to filter out non-text areas. The stroke width map (SWM) is proposed to compute the stroke width by finding the maximal distance among the shortest distances with the same nearest boundary pixel. It can reflect the true stroke width and adapt to complex situations by improving the accuracy of stroke width. It does not require any training process. Experiment results demonstrate that this approach can achieve a good performance of text extraction with various degradations, highlighting the system efficiency and the text extraction capability.