Automatic Text Detection in Scene Images

K. Vaidehi, S. Bhuvaneswari

Abstract: Automatically detecting and extracting text from the digital images present many challenging research issues. The proposed system detects text automatically using connected component labeling method. In this work, first, color scene images are taken as input image which is converted into gray scale image, then Otsu thresholding algorithm is used to binarize the gray scale image. To detect the connected regions, connected component labeling method is used and to retain the text regions alone, set of selection/rejection criteria is used.

Keywords: Text Detection, Otsu?s thresholding, connected component labeling, Morphological operations.