String Extraction from Scene Images Using Color Information

In this paper, a robust character string extraction method from scene images using both color and luminance information is presented. The method makes good binary images which separate characters and other objects and extracts character strings using character connectivity. In binarization using color information, color-clustering classifies colors to some representative colors, and object-clustering classifies the representative colors. In binarization using luminance information, luminance image is divided into small regions, and a threshold to binarize is determined for each pixel in each region using only its surrounding region. Finally, in each binary image, a character string is extracted using character connectivity.

[1]  Hon-Son Don,et al.  A noise attribute thresholding method for document image binarization , 1995, Proceedings of 3rd International Conference on Document Analysis and Recognition.

[2]  Kongqiao Wang,et al.  Character location in scene images from digital camera , 2003, Pattern Recognit..