A Novel Algorithm for Text Localization in Color Images
LI Jun-shan · Radio and communications technology · 2008
Texts in color images contain important semantic information for image indexing and retrieval.A novel connected component based algorithm for text localization in color images is proposed in this paper.Firstly,color image is multi-channel decomposed into binary images of representative colors.Connected component labeling is performed in each binary image,and properties of each connected component are calculated.And then according to apriority knowledge of properties of connected components,non-text connected components are eliminated to get candidate text connected components.Finally,alignment analysis is used to localizing the texts from the candidate text connected components in color images.Experimental results show the proposed algorithm can localize text with faster speed and better accuracy compared with other method tested.