The popularity of digital devices like cameras, smart phones and embedded visual processing is increasing. The detection of text from natural scene images has gained much attention of researchers because of its various applications in our daily lives. Also due to the presence of useful information about scene in the image, it is being studied by many researchers. The text detection is difficult as text present in natural scene images exhibit different characteristics with respect to size, font, color and orientation. In this paper different categories of text detection methods are discussed. The classification of existing methods incorporates edge based, connected component based, texture based, stroke based and hybrid techniques. It also exposes the comparison between existing methods of text detection.