A modified SWT based text-image separation in natural scene images

摘要

Automatic text detection from natural scene images catches everyone's attention because of applications in robotic navigation, aid for visually impaired people, image retrieval, number plate detection, content-based web search, intelligent transport system etc. In this paper, our aim is to detect text in natural images and textual images using stroke-width transform method. An image operator (SWT) which finds the stroke-width of every pixel has been used. Experimental results indicate that the features of SWT make it reliable and robust to detect text independent of scale, direction, background and font. This method can be categorized as a connected component based approach depending upon the way in which possible text areas are detected.